A Wearable Textile 3D Gesture Recognition Sensor Based on Screen-Printing Technology

Research has developed various solutions in order for computers to recognize hand gestures in the context of human machine interface (HMI). The design of a successful hand gesture recognition system must address functionality and usability. The gesture recognition market has evolved from touchpads to touchless sensors, which do not need direct contact. Their application in textiles ranges from the field of medical environments to smart home applications and the automotive industry. In this paper, a textile capacitive touchless sensor has been developed by using screen-printing technology. Two different designs were developed to obtain the best configuration, obtaining good results in both cases. Finally, as a real application, a complete solution of the sensor with wireless communications is presented to be used as an interface for a mobile phone.


Introduction
The ability of computers to recognize hand gestures is essential for progress in human-computer interaction (HCI) and human machine interaction (HMI) [1]. Gesture recognition enables humans to interact with computers and machines without any physical contact in a more natural way. It is an alternative method for interacting with computers without using traditional peripheral devices, such as the keyboard or the mouse. One of the most relevant devices is the Microsoft Kinect sensor, now updated as Azure Kinect DK sensor, both consisting of advanced sensing hardware combining a VGA (Video Graphics Array) video camera, an infrared depth sensor and a multi-array microphone allowing people to interact with the games using their body [2]. In addition, this device includes a complex microprocessor that is able to run advanced machine-learning algorithms on a graphics processing unit (GPU) in parallel, tracking simultaneously up to six people. Although face and other body parts, considered as full body interaction [3], can be used to interact with computers, hand gesture recognition is the most popular solution for different reasons. Factors such as end-user application, reliability, cost, and context influence the choice of one technology in detriment of another.
Hand gesture recognition based human machine interface (HMI) is an attractive topic because it is especially important in both industrial and domestic applications, to evolve the way we interact with our environment. Hand gesture recognition allows us to control elements in a 3D space, rather than traditional interfaces, such as the mouse or the keyboard, that are limited to a 2D environment. These features have an important impact on CAD (Computer-Aided Design) applications, 3D gaming, or Virtual Reality. Hence, gesture recognition has a huge application ranging from very simple  The design suggested by Microchip [40] consists of a sensor made in rigid or double-sided flexible PCB (Printed Circuit Board), where the Rx electrodes are distributed on the four cardinal points of the upper face and on the centre. The Tx electrode is located on the underside. This design is used when the operation does not need a battery, and the electrical noise is low. When the external noise is high or the functioning uses a battery, three layers are needed due to the addition of a ground plane. In the PCB designs, four conductive layers are implemented, leaving one of them with no use. In the case of utilizing a four layer design, the upper layer is used for Rx electrodes, the second internal layer for the Tx signal and the underside layer for the ground plane ( Figure 3).  The design suggested by Microchip [40] consists of a sensor made in rigid or double-sided flexible PCB (Printed Circuit Board), where the Rx electrodes are distributed on the four cardinal points of the upper face and on the centre. The Tx electrode is located on the underside. This design is used when the operation does not need a battery, and the electrical noise is low. When the external noise is high or the functioning uses a battery, three layers are needed due to the addition of a ground plane. In the PCB designs, four conductive layers are implemented, leaving one of them with no use. In the case of utilizing a four layer design, the upper layer is used for Rx electrodes, the second internal layer for the Tx signal and the underside layer for the ground plane ( Figure 3). The design suggested by Microchip [40] consists of a sensor made in rigid or double-sided flexible PCB (Printed Circuit Board), where the Rx electrodes are distributed on the four cardinal points of the upper face and on the centre. The Tx electrode is located on the underside. This design is used when the operation does not need a battery, and the electrical noise is low. When the external noise is high or the functioning uses a battery, three layers are needed due to the addition of a ground plane. In the PCB designs, four conductive layers are implemented, leaving one of them with no use. In the case of utilizing a four layer design, the upper layer is used for Rx electrodes, the second internal layer for the Tx signal and the underside layer for the ground plane ( Figure 3).  The Rx and Tx layers must be of conductive material such as copper but can be of indium-tin oxide (ITO) or similar. The electrode isolation can be made of any non-conductive material (FR4, glass, PET, etc.). Microchip proposes two designs [40] for MGC3130, of different and compatible sensors: 1. Standard sensor (Tx signal amplitude of 2.85 V). Useful for small or medium-sized devices and mandatory for devices with a weak connection to ground, that is, with battery. 2. Booster sensor (Tx signal amplitude between 5 and 18 V) allowing bigger sensors and recognition ranges. Figure 4 shows the block diagram of the MGC3130 controller. It consists of the analogue frontend unit connected to the Tx and Rx electrodes, the signal processing unit that receives data from the front-end and is assisted by a GestIC library and, finally, a communication interface for the data interconnection between MGC3130 and another microcontroller.
The MGC3130 controller is parameterizable, that is, it can be reconfigured for each type of sensor and application. Thus, the most important part of the design of this system is the sensor characterization. It can be used to detect hand movements or only finger movements, influencing in the size of the sensor. This research uses the hand, which can be used in parallel, perpendicularly or with a certain angle to the sensitive area. The sensitive area is always defined inside of the four cardinal Rx electrodes. The geometrical form of the sensitive area is a key factor. This depends on the electrode geometry but also on the electrode location. For the design of the sensor, it is necessary to have some knowledge of its electrical parameters (shown in Figure 5). This electrical model allows us to evaluate the characteristics of the system and points out the dependencies between the electrodes, the MGC3130 controller and the hand. In particular, the capacitances of the electrodes which are coupled to the input/output sections of the MGC3130 integrated circuit. The Rx and Tx layers must be of conductive material such as copper but can be of indium-tin oxide (ITO) or similar. The electrode isolation can be made of any non-conductive material (FR4, glass, PET, etc.). Microchip proposes two designs [40] for MGC3130, of different and compatible sensors: Standard sensor (Tx signal amplitude of 2.85 V). Useful for small or medium-sized devices and mandatory for devices with a weak connection to ground, that is, with battery.

2.
Booster sensor (Tx signal amplitude between 5 and 18 V) allowing bigger sensors and recognition ranges. Figure 4 shows the block diagram of the MGC3130 controller. It consists of the analogue front-end unit connected to the Tx and Rx electrodes, the signal processing unit that receives data from the front-end and is assisted by a GestIC library and, finally, a communication interface for the data interconnection between MGC3130 and another microcontroller.
The MGC3130 controller is parameterizable, that is, it can be reconfigured for each type of sensor and application. Thus, the most important part of the design of this system is the sensor characterization. It can be used to detect hand movements or only finger movements, influencing in the size of the sensor. This research uses the hand, which can be used in parallel, perpendicularly or with a certain angle to the sensitive area. The sensitive area is always defined inside of the four cardinal Rx electrodes. The geometrical form of the sensitive area is a key factor. This depends on the electrode geometry but also on the electrode location. For the design of the sensor, it is necessary to have some knowledge of its electrical parameters (shown in Figure 5). This electrical model allows us to evaluate the characteristics of the system and points out the dependencies between the electrodes, the MGC3130 controller and the hand. In particular, the capacitances of the electrodes which are coupled to the input/output sections of the MGC3130 integrated circuit.
The output signal, VTx, is found on the Tx pin of the MGC3130 integrated circuit, whose capacitance to ground is CTxG. eTx and eRx represent the transmission and reception electrodes. The parameterization of the corresponding circuit is based on the CRxTx, CRxG, and CL capacitances. CRxTx is the capacitance between the Tx and Rx electrodes, CRxG is the capacitance between the Rx electrode and ground and CL is the coupled capacitance between the transmission Tx pin and the Rx reception line. Finally, CH is the capacitance between the hand and the Rx electrode. CH is represented as a variable capacitor, since the corresponding capacitance depends on the hand and its position regarding the Rx electrodes. The eRx reception electrode measures the electric field potential. When a conductive object, like a hand, interacts with the electric field, the CH capacitance of the reception electrode changes in the range of femtofarads and the corresponding variation is detected by the MGC3130 integrated circuit. The embedded table in Figure 5 shows the typical range of these capacitances.

Microchip Sensor Design
The sensor can have different sizes and shapes, such as square, circular, oval, or rectangular, as long as the 1:3 length to width ratio is not exceeded. Microchip recommends a maximum of 140 × 140 mm (or a 140 mm diameter) and a minimum of 20 × 20 mm (diameter of 20 mm). The sensor must have a minimum of two layers but can be made with three layers: Rx on the upper layer, Tx on the bottom layer and, if necessary, GND as a third layer [40].
The output signal, VTx, is found on the Tx pin of the MGC3130 integrated circuit, whose capacitance to ground is CTxG. eTx and eRx represent the transmission and reception electrodes. The parameterization of the corresponding circuit is based on the CRxTx, CRxG, and CL capacitances. CRxTx is the capacitance between the Tx and Rx electrodes, CRxG is the capacitance between the Rx electrode and ground and CL is the coupled capacitance between the transmission Tx pin and the Rx reception line. Finally, CH is the capacitance between the hand and the Rx electrode. CH is represented as a variable capacitor, since the corresponding capacitance depends on the hand and its position regarding the Rx electrodes. The eRx reception electrode measures the electric field potential. When a conductive object, like a hand, interacts with the electric field, the CH capacitance of the reception electrode changes in the range of femtofarads and the corresponding variation is detected by the MGC3130 integrated circuit. The embedded table in Figure 5 shows the typical range of these capacitances.

Microchip Sensor Design
The sensor can have different sizes and shapes, such as square, circular, oval, or rectangular, as long as the 1:3 length to width ratio is not exceeded. Microchip recommends a maximum of 140 × 140 mm (or a 140 mm diameter) and a minimum of 20 × 20 mm (diameter of 20 mm). The sensor must have a minimum of two layers but can be made with three layers: Rx on the upper layer, Tx on the bottom layer and, if necessary, GND as a third layer [40].

Microchip Sensor Design
The sensor can have different sizes and shapes, such as square, circular, oval, or rectangular, as long as the 1:3 length to width ratio is not exceeded. Microchip recommends a maximum of 140 × 140 mm (or a 140 mm diameter) and a minimum of 20 × 20 mm (diameter of 20 mm). The sensor must have a minimum of two layers but can be made with three layers: Rx on the upper layer, Tx on the bottom layer and, if necessary, GND as a third layer [40]. Figure 6 shows a typical design. On the upper layer, the four perimeter electrodes can be observed, as well as the optional central electrode, whereas the Tx electrode is located on the bottom layer. The size Sensors 2019, 19, 5068 7 of 32 of the Rx electrodes determines the sensitive area. These Rx electrodes must always remain delimiting the aforementioned sensitive area to obtain the maximum resolution for the x, y, and z coordinates.  Figure 6 shows a typical design. On the upper layer, the four perimeter electrodes can be observed, as well as the optional central electrode, whereas the Tx electrode is located on the bottom layer. The size of the Rx electrodes determines the sensitive area. These Rx electrodes must always remain delimiting the aforementioned sensitive area to obtain the maximum resolution for the x, y, and z coordinates. With the aim of obtaining a high sensitivity in the sensitive area, the area of the reception electrodes and the hand should be of the same order of magnitude. Hence, the four Rx electrodes of the cardinal points should have a lengthened shape to increment the coupling between the Rx electrodes and the hand. If the recognition range must be symmetric in both directions, the electrode design must be symmetric. In any case, the length of both vertical and horizontal electrodes must be balanced. The recommended distance between the Rx electrodes is of 1 to 5 mm. The most influential factor of the sensitivity to the hand is the width of the perimeter Rx electrodes. The bigger the width of the electrode is, the higher the sensitivity is. The receiver signal sensitivity measures the influence of the hand capacitance when a hand approaches the system. As a consequence, the Rx signal changes in presence of a hand, deviating from a reference value. The value of the deviation is measured and named Signal Deviation. It is the basis for the recognition of gestures [41]. Microchip references the signal deviation obtained in MGC3130 with respect to the distance of the hand compared with the width of the Rx electrodes ( Figure 7). Logically, the larger the distance of the hand on the sensitive area is, the less the signal deviation is. The width of the electrode also influences the signal deviation. The wider the Rx electrode is, the greater the signal deviation is. However, an increase in the area of the Rx electrode implies a limitation in the range of gesture recognition. A best practice is to choose an Rx electrode width between 4% and 7% of the length. Moreover, an overly large Rx electrode causes higher capacitance between Rx and Tx and also between Rx and the ground. As a result of that, a loss of signal sensitivity is produced. With the aim of obtaining a high sensitivity in the sensitive area, the area of the reception electrodes and the hand should be of the same order of magnitude. Hence, the four Rx electrodes of the cardinal points should have a lengthened shape to increment the coupling between the Rx electrodes and the hand. If the recognition range must be symmetric in both directions, the electrode design must be symmetric. In any case, the length of both vertical and horizontal electrodes must be balanced. The recommended distance between the Rx electrodes is of 1 to 5 mm. The most influential factor of the sensitivity to the hand is the width of the perimeter Rx electrodes. The bigger the width of the electrode is, the higher the sensitivity is. The receiver signal sensitivity measures the influence of the hand capacitance when a hand approaches the system. As a consequence, the Rx signal changes in presence of a hand, deviating from a reference value. The value of the deviation is measured and named Signal Deviation. It is the basis for the recognition of gestures [41]. Microchip references the signal deviation obtained in MGC3130 with respect to the distance of the hand compared with the width of the Rx electrodes ( Figure 7). Logically, the larger the distance of the hand on the sensitive area is, the less the signal deviation is. The width of the electrode also influences the signal deviation. The wider the Rx electrode is, the greater the signal deviation is. However, an increase in the area of the Rx electrode implies a limitation in the range of gesture recognition. A best practice is to choose an Rx electrode width between 4% and 7% of the length. Moreover, an overly large Rx electrode causes higher capacitance between Rx and Tx and also between Rx and the ground. As a result of that, a loss of signal sensitivity is produced. In the case of the central Rx electrode, a meshed design is recommended. The mesh design increases the sensitivity due to the reduction of the effective area of the Rx electrode and to the coupling to the Tx electrode. An acceptable value for the mesh would be around 5% and 20%.
The CRxGND capacitance must be as low as possible. That is, the distance between the Rx electrodes and ground must be as high as possible.
Regarding the Tx electrode, it must be located under the Rx electrodes and must have a shape and size to cover all the reception area. On the other hand, the Tx electrode must have a low coupling with the Rx electrodes (CRxTx) and with ground (CTxG). To reduce the CRxTx capacitance, the distance between the Tx and the Rx electrodes must be as long as possible. The optimum distance between these two layers (t) will depend on the relative permittivity (εr) of the isolation material between both layers. Microchip recommends a t > εr/5, hence, for a FR4 material (glass-reinforced epoxy laminate material εr = 5), the thickness can be of 1 mm, for plastic (εr = 3), it can be of 0.6 mm and, for glass (εr = 6), it can be of 1.2 mm. Microchip recommends a minimum value between 1 to 2.5 mm, a value of 50% to 100% of the Rx electrode width being considered optimal. In any case, the calculated thickness will be the minimum one, achieving a better response as this distance increases.
Regarding the CTxGND capacitance, it cannot be greater than the conductive capacitance of MGC3130 that is 1 nF. To reduce it, the Tx electrode must have a large surface and a meshed design of between 20% and 50% of the surface. When the CTxGND capacitance is not low enough, an operational amplifier (op-amp) must be inserted between the Tx pin and the Tx electrode ( Figure 8). Lastly, Microchip recommends an overlap between the Tx electrode and the external edges of the Rx electrodes greater than 3 mm (Figure 9).  In the case of the central Rx electrode, a meshed design is recommended. The mesh design increases the sensitivity due to the reduction of the effective area of the Rx electrode and to the coupling to the Tx electrode. An acceptable value for the mesh would be around 5% and 20%.
The C RxGND capacitance must be as low as possible. That is, the distance between the Rx electrodes and ground must be as high as possible.
Regarding the Tx electrode, it must be located under the Rx electrodes and must have a shape and size to cover all the reception area. On the other hand, the Tx electrode must have a low coupling with the Rx electrodes (C RxTx ) and with ground (C TxG ). To reduce the C RxTx capacitance, the distance between the Tx and the Rx electrodes must be as long as possible. The optimum distance between these two layers (t) will depend on the relative permittivity (ε r ) of the isolation material between both layers. Microchip recommends a t > ε r /5, hence, for a FR4 material (glass-reinforced epoxy laminate material ε r = 5), the thickness can be of 1 mm, for plastic (ε r = 3), it can be of 0.6 mm and, for glass (ε r = 6), it can be of 1.2 mm. Microchip recommends a minimum value between 1 to 2.5 mm, a value of 50% to 100% of the Rx electrode width being considered optimal. In any case, the calculated thickness will be the minimum one, achieving a better response as this distance increases.
Regarding the C TxGND capacitance, it cannot be greater than the conductive capacitance of MGC3130 that is 1 nF. To reduce it, the Tx electrode must have a large surface and a meshed design of between 20% and 50% of the surface. When the C TxGND capacitance is not low enough, an operational amplifier (op-amp) must be inserted between the Tx pin and the Tx electrode ( Figure 8). Lastly, Microchip recommends an overlap between the Tx electrode and the external edges of the Rx electrodes greater than 3 mm (Figure 9). In the case of the central Rx electrode, a meshed design is recommended. The mesh design increases the sensitivity due to the reduction of the effective area of the Rx electrode and to the coupling to the Tx electrode. An acceptable value for the mesh would be around 5% and 20%.
The CRxGND capacitance must be as low as possible. That is, the distance between the Rx electrodes and ground must be as high as possible.
Regarding the Tx electrode, it must be located under the Rx electrodes and must have a shape and size to cover all the reception area. On the other hand, the Tx electrode must have a low coupling with the Rx electrodes (CRxTx) and with ground (CTxG). To reduce the CRxTx capacitance, the distance between the Tx and the Rx electrodes must be as long as possible. The optimum distance between these two layers (t) will depend on the relative permittivity (εr) of the isolation material between both layers. Microchip recommends a t > εr/5, hence, for a FR4 material (glass-reinforced epoxy laminate material εr = 5), the thickness can be of 1 mm, for plastic (εr = 3), it can be of 0.6 mm and, for glass (εr = 6), it can be of 1.2 mm. Microchip recommends a minimum value between 1 to 2.5 mm, a value of 50% to 100% of the Rx electrode width being considered optimal. In any case, the calculated thickness will be the minimum one, achieving a better response as this distance increases.
Regarding the CTxGND capacitance, it cannot be greater than the conductive capacitance of MGC3130 that is 1 nF. To reduce it, the Tx electrode must have a large surface and a meshed design of between 20% and 50% of the surface. When the CTxGND capacitance is not low enough, an operational amplifier (op-amp) must be inserted between the Tx pin and the Tx electrode ( Figure 8). Lastly, Microchip recommends an overlap between the Tx electrode and the external edges of the Rx electrodes greater than 3 mm (Figure 9).  Apart from the Rx and Tx layers, a third layer, GND, must be added in case of systems with no battery in high noise environments or in case of battery operation. When the system operates with batteries, this GND layer is mandatory but, in systems with ground connection, this layer is optional. In any case, the GND layer confers stability and noise immunity but at the expense of losing sensitivity, between 10% and 20%. As aforementioned, the CTxGND capacitance must be lower than 1 nF. To achieve this, the thickness between the Tx and GND layers must be increased using materials with a greater relative permittivity and mesh design.
The Rx and Tx electrodes are not only limited to the sensitive area of the sensor but are also formed by the conductive lines that join these electrodes with the MGC3130 controller. Hence, these conductive lines influence the gesture detection as well. For this reason, these lines must be designed with as shorter length as possible, inside of the sensitive area and, if possible, with a distance from the Tx electrode larger than 0.15 mm.
Before proceeding with the design of the sensor with a textile, a study of the sensor of Microchip was performed. The reference sensor had a sensitive area of 95 x 60 mm with a size of 120 x 85 mm. This study helped with the validation of the results obtained with the textile version. The sensor characteristics are shown in Figure 10.
The PCB (Printed Circuit Board) follows a four layers design. The Rx electrodes are located on the upper layer, the second layer (internal) is not used, the third layer (internal) contains the Tx electrode, and the bottom layer includes the GND plane. The dielectric between the conductive layers is of FR4 material. The cross distribution of the PCB is shown in Figure 11. The relative permittivity (εr) of FR4 is considered to be 4.8. Apart from the Rx and Tx layers, a third layer, GND, must be added in case of systems with no battery in high noise environments or in case of battery operation. When the system operates with batteries, this GND layer is mandatory but, in systems with ground connection, this layer is optional. In any case, the GND layer confers stability and noise immunity but at the expense of losing sensitivity, between 10% and 20%. As aforementioned, the C TxGND capacitance must be lower than 1 nF. To achieve this, the thickness between the Tx and GND layers must be increased using materials with a greater relative permittivity and mesh design.
The Rx and Tx electrodes are not only limited to the sensitive area of the sensor but are also formed by the conductive lines that join these electrodes with the MGC3130 controller. Hence, these conductive lines influence the gesture detection as well. For this reason, these lines must be designed with as shorter length as possible, inside of the sensitive area and, if possible, with a distance from the Tx electrode larger than 0.15 mm.
Before proceeding with the design of the sensor with a textile, a study of the sensor of Microchip was performed. The reference sensor had a sensitive area of 95 × 60 mm with a size of 120 × 85 mm. This study helped with the validation of the results obtained with the textile version. The sensor characteristics are shown in Figure 10.
The PCB (Printed Circuit Board) follows a four layers design. The Rx electrodes are located on the upper layer, the second layer (internal) is not used, the third layer (internal) contains the Tx electrode, and the bottom layer includes the GND plane. The dielectric between the conductive layers is of FR4 material. The cross distribution of the PCB is shown in Figure 11. The relative permittivity (ε r ) of FR4 is considered to be 4.8.   Table 1 shows the different values obtained for CTxRx. The range acceptable by Microchip for this capacitance oscillates between 5 and 30 pF. The value for this PCB provided by Microchip is shown as well. Note the higher value corresponding to the central electrode due to its larger area.  Table 2 shows the different values obtained for CTxGND. The value for this capacitance acceptable by Microchip is lower than 1 nF.   Table 1 shows the different values obtained for CTxRx. The range acceptable by Microchip for this capacitance oscillates between 5 and 30 pF. The value for this PCB provided by Microchip is shown as well. Note the higher value corresponding to the central electrode due to its larger area.  Table 2 shows the different values obtained for CTxGND. The value for this capacitance acceptable by Microchip is lower than 1 nF.  Table 1 shows the different values obtained for C TxRx . The range acceptable by Microchip for this capacitance oscillates between 5 and 30 pF. The value for this PCB provided by Microchip is shown as well. Note the higher value corresponding to the central electrode due to its larger area.  Table 2 shows the different values obtained for C TxGND . The value for this capacitance acceptable by Microchip is lower than 1 nF.  Table 3 shows the different values obtained for C RxGND . The range for this capacitance acceptable by Microchip oscillates between 5 and 30 pF.  Figure 12a shows the Tx signal generated by MGC3130, in this case of frequency 115 kHz and amplitude 2.84 V. Figure 12b shows the signal received from one of the Rx electrodes when there is no object on the sensitive area modifying the electrical field. As expected, the frequency is 115 kHz, but the amplitude of the received signal is 1.82 V. Figure 12c shows the same Rx signal, but introducing an object in the sensitive area. The amplitude of the signal varies up to 2.18 V.   Table 3 shows the different values obtained for CRxGND. The range for this capacitance acceptable by Microchip oscillates between 5 and 30 pF.  Figure 12a shows the Tx signal generated by MGC3130, in this case of frequency 115 kHz and amplitude 2.84 V. Figure 12b shows the signal received from one of the Rx electrodes when there is no object on the sensitive area modifying the electrical field. As expected, the frequency is 115 kHz, but the amplitude of the received signal is 1.82 V. Figure 12c shows the same Rx signal, but introducing an object in the sensitive area. The amplitude of the signal varies up to 2.18 V.

Textile 3D Gesture Sensor Design
Starting from the design data for the sensitive electrodes from Microchip, several of our own designs have been developed based on textile substrates. A similar design to Microchip PCB has been made but using the textile substrate as the dielectric layer.
The manufacturing technology used to implement this type of sensor was based on serigraphic technology of thick film. The screen-printing process consists of forcing pastes of different characteristics over a substrate through some screens using squeegees. Openings in the screen define the pattern that will be printed on the substrate by serigraphy. The final thickness of the pastes can be adjusted by varying the thickness of the screens. In the same way as for a PCB manufacturing, conductive materials and dielectric materials have been employed. Conductive silver ink has been employed to make the electrodes. In the case of the dielectrics, textile, dielectric inks, and polyurethane plastic films have been the materials used.
Different types of textiles have been selected in function of the material used for their manufacturing. The aim is to study their relative permittivity as well as their thickness. Both parameters influence in the value of the associated capacitances as analysed previously.
For each design, type of textile, and inks, measures of their physical and electrical parameters have been taken. The C TxRx , C RxGND and C TxGND capacitances have been measured as well.
The first design made, named 3DS-1, is shown in Figure 13. It consists of a ground plane layer (Figure 13a), a layer containing the Tx electrode and the connection lines of Rx with MGC3130 ( Figure 13b), a dielectric layer provided with vias that allow the connection between the Rx layer and the Rx connection lines (Figure 13c), and the Rx electrode layer with 4 perimeter electrodes and one central ( Figure 13d). The dimensions of the sensor are identical to the 95 × 60 sensor from Microchip explained previously. In this design, the connection lines of the Rx electrodes have been placed on the transmit layer to minimize interferences.

Textile 3D Gesture Sensor Design
Starting from the design data for the sensitive electrodes from Microchip, several of our own designs have been developed based on textile substrates. A similar design to Microchip PCB has been made but using the textile substrate as the dielectric layer.
The manufacturing technology used to implement this type of sensor was based on serigraphic technology of thick film. The screen-printing process consists of forcing pastes of different characteristics over a substrate through some screens using squeegees. Openings in the screen define the pattern that will be printed on the substrate by serigraphy. The final thickness of the pastes can be adjusted by varying the thickness of the screens. In the same way as for a PCB manufacturing, conductive materials and dielectric materials have been employed. Conductive silver ink has been employed to make the electrodes. In the case of the dielectrics, textile, dielectric inks, and polyurethane plastic films have been the materials used.
Different types of textiles have been selected in function of the material used for their manufacturing. The aim is to study their relative permittivity as well as their thickness. Both parameters influence in the value of the associated capacitances as analysed previously.
For each design, type of textile, and inks, measures of their physical and electrical parameters have been taken. The CTxRx, CRxGND and CTxGND capacitances have been measured as well.
The first design made, named 3DS-1, is shown in Figure 13. It consists of a ground plane layer (Figure 13a), a layer containing the Tx electrode and the connection lines of Rx with MGC3130 ( Figure  13b), a dielectric layer provided with vias that allow the connection between the Rx layer and the Rx connection lines (Figure 13c), and the Rx electrode layer with 4 perimeter electrodes and one central ( Figure 13d). The dimensions of the sensor are identical to the 95 × 60 sensor from Microchip explained previously. In this design, the connection lines of the Rx electrodes have been placed on the transmit layer to minimize interferences.  Figure 14 shows the cross-section of the sensor. The complete structure contains 5 layers since the textile substrate is used as a dielectric layer between the ground plane layer and the Tx electrode layer. The different layers, except the substrate, are screen-printed. Silver ink was employed for the  Figure 14 shows the cross-section of the sensor. The complete structure contains 5 layers since the textile substrate is used as a dielectric layer between the ground plane layer and the Tx electrode layer. The different layers, except the substrate, are screen-printed. Silver ink was employed for the conductive materials. Dielectric inks were used in the case of the dielectric layer between the Rx electrode layer and the Tx electrode layer. With this design, high capacitances are obtained out of the range recommended by Microchip for use with MGC3130. In part, this is due to the solution employed for the dielectric with worse performance than expected. This will be discussed in the Results Section. conductive materials. Dielectric inks were used in the case of the dielectric layer between the Rx electrode layer and the Tx electrode layer. With this design, high capacitances are obtained out of the range recommended by Microchip for use with MGC3130. In part, this is due to the solution employed for the dielectric with worse performance than expected. This will be discussed in the Results Section.  Figure 13 the textile substrate between the ground plane layer and the Tx electrode layer can be observed.
The second design made, named 3DS-2, is shown in Figure 15. It consists of a ground plane layer (Figure 15a), a dielectric layer between the ground plane layer and the Tx layer (Figure 15b), a layer containing the transmission Tx electrode Figure 15c) and a layer with the reception Rx electrodes (Figure 15d). The textile substrate acts as a dielectric between Rx and Tx layers. In this case, making the vias on the textile substrate is not viable. Hence, the connection lines between the Rx electrodes and MGC3130 have been made on the same Rx layer. This design allows us to decrease the CTxRx and CRxGND capacitances since it is possible to utilize a textile substrate with a high thickness.  Figure 16 shows the cross-section of the sensor. The complete structure contains 5 layers since the textile substrate is used as a dielectric layer between the Rx electrode layer and the Tx electrode layer. The different layers, except the substrate, are screen-printed employing silver ink for the conductive materials and dielectric inks in the case of the dielectric layer between the Rx electrode layer and the Tx electrode layer. The main problem of this design comes from the alignment of the  Figure 13 the textile substrate between the ground plane layer and the Tx electrode layer can be observed.
The second design made, named 3DS-2, is shown in Figure 15. It consists of a ground plane layer (Figure 15a), a dielectric layer between the ground plane layer and the Tx layer (Figure 15b), a layer containing the transmission Tx electrode Figure 15c) and a layer with the reception Rx electrodes (Figure 15d). The textile substrate acts as a dielectric between Rx and Tx layers. In this case, making the vias on the textile substrate is not viable. Hence, the connection lines between the Rx electrodes and MGC3130 have been made on the same Rx layer. This design allows us to decrease the C TxRx and C RxGND capacitances since it is possible to utilize a textile substrate with a high thickness. conductive materials. Dielectric inks were used in the case of the dielectric layer between the Rx electrode layer and the Tx electrode layer. With this design, high capacitances are obtained out of the range recommended by Microchip for use with MGC3130. In part, this is due to the solution employed for the dielectric with worse performance than expected. This will be discussed in the Results Section. Figure 14. Cross-section of the 3DS-1 sensor. In addition to the 4 layers shown in Figure 13 the textile substrate between the ground plane layer and the Tx electrode layer can be observed.
The second design made, named 3DS-2, is shown in Figure 15. It consists of a ground plane layer (Figure 15a), a dielectric layer between the ground plane layer and the Tx layer (Figure 15b), a layer containing the transmission Tx electrode Figure 15c) and a layer with the reception Rx electrodes (Figure 15d). The textile substrate acts as a dielectric between Rx and Tx layers. In this case, making the vias on the textile substrate is not viable. Hence, the connection lines between the Rx electrodes and MGC3130 have been made on the same Rx layer. This design allows us to decrease the CTxRx and CRxGND capacitances since it is possible to utilize a textile substrate with a high thickness.  Figure 16 shows the cross-section of the sensor. The complete structure contains 5 layers since the textile substrate is used as a dielectric layer between the Rx electrode layer and the Tx electrode layer. The different layers, except the substrate, are screen-printed employing silver ink for the conductive materials and dielectric inks in the case of the dielectric layer between the Rx electrode layer and the Tx electrode layer. The main problem of this design comes from the alignment of the  Figure 16 shows the cross-section of the sensor. The complete structure contains 5 layers since the textile substrate is used as a dielectric layer between the Rx electrode layer and the Tx electrode layer. The different layers, except the substrate, are screen-printed employing silver ink for the conductive materials and dielectric inks in the case of the dielectric layer between the Rx electrode layer and the Tx electrode layer. The main problem of this design comes from the alignment of the connection lines. The use of a textile substrate makes it difficult for the correct alignment of the Rx and Tx layers. For this reason, a final design with a different connector position was proposed to solve this problem. The design is shown in Figure 16. connection lines. The use of a textile substrate makes it difficult for the correct alignment of the Rx and Tx layers. For this reason, a final design with a different connector position was proposed to solve this problem. The design is shown in Figure 16.  Figure 15, the textile substrate between the Rx layer and the Tx electrode layer can be observed.
Both designs have been developed with 8 different types of textiles (Tables 4 and 5). The aim is to determine the influence of the thickness and the relative permittivity of the textiles on the sensor. Each one of the employed textiles have been characterized physically and electrically.
Regarding the inks, one conductive silver ink ( Table 6) and three types of dielectric inks have been employed ( Table 7). The inks influence noticeably the final capacitance at a thickness level as well as at a relative permittivity level. Heat sealed polyurethanes (Table 8) have been employed as well in some sensors as substitutes of the dielectric inks.
Once the different sensors were built, they have been characterized electrically by means of the determination of their CTxRx, CRxGND, and CTxGND capacitances. Lastly, the sensitivity of the different sensors has been determined using a graphical user interface software from Microchip, AUREA.

Materials
Different cotton, polyester and mixed fabrics with different fabric densities, yarn diameter, and weave have been used. Tables 4 and 5 show the main characteristics of the fabrics used.
The conductive ink (Table 6) used is silver IPC-603X from INKRON (Kutojantie, Espoo, Finland). Three dielectric inks (Table 7)  Two polyurethane films (Table 8) have been used. The films are EU94DS from DELSTARD and UAF445 from ADHESIVE FILMS.  Figure 16. Cross-section of the 3DS-2 sensor. In addition to the 4 layers shown in Figure 15, the textile substrate between the Rx layer and the Tx electrode layer can be observed.
Both designs have been developed with 8 different types of textiles (Tables 4 and 5). The aim is to determine the influence of the thickness and the relative permittivity of the textiles on the sensor. Each one of the employed textiles have been characterized physically and electrically. determination of their CTxRx, CRxGND, and CTxGND capacitances. Lastly, the sensitivity of the different sensors has been determined using a graphical user interface software from Microchip, AUREA.

Materials
Different cotton, polyester and mixed fabrics with different fabric densities, yarn diameter, and weave have been used. Tables 4 and 5 show the main characteristics of the fabrics used.
Two polyurethane films (Table 8) have been used. The films are EU94DS from DELSTARD and UAF445 from ADHESIVE FILMS. Once the different sensors were built, they have been characterized electrically by means of the determination of their CTxRx, CRxGND, and CTxGND capacitances. Lastly, the sensitivity of the different sensors has been determined using a graphical user interface software from Microchip, AUREA.

Materials
Different cotton, polyester and mixed fabrics with different fabric densities, yarn diameter, and weave have been used. Tables 4 and 5 show the main characteristics of the fabrics used.
Two polyurethane films (Table 8) have been used. The films are EU94DS from DELSTARD and UAF445 from ADHESIVE FILMS. determination of their CTxRx, CRxGND, and CTxGND capacitances. Lastly, the sensitivity of the different sensors has been determined using a graphical user interface software from Microchip, AUREA.

Materials
Different cotton, polyester and mixed fabrics with different fabric densities, yarn diameter, and weave have been used. Tables 4 and 5 show the main characteristics of the fabrics used.
The conductive ink (Table 6) used is silver IPC-603X from INKRON (Kutojantie, Espoo, Finland). Three dielectric inks (Table 7) with different flexibility and stretchability characteristics have been used. The inks are 127-48D from CREATIVE (Lamesa, Texas, USA), DI-7542 from EMS (Delaware, Ohio, USA) and IPD-670 from INKRON. Two polyurethane films (Table 8) have been used. The films are EU94DS from DELSTARD and UAF445 from ADHESIVE FILMS. determination of their CTxRx, CRxGND, and CTxGND capacitances. Lastly, the sensitivity of the different sensors has been determined using a graphical user interface software from Microchip, AUREA.

Materials
Different cotton, polyester and mixed fabrics with different fabric densities, yarn diameter, and weave have been used. Tables 4 and 5 show the main characteristics of the fabrics used.
The conductive ink (Table 6) used is silver IPC-603X from INKRON (Kutojantie, Espoo, Finland). Three dielectric inks (Table 7) with different flexibility and stretchability characteristics have been used. The inks are 127-48D from CREATIVE (Lamesa, Texas, USA), DI-7542 from EMS (Delaware, Ohio, USA) and IPD-670 from INKRON. Two polyurethane films (Table 8) have been used. The films are EU94DS from DELSTARD and UAF445 from ADHESIVE FILMS.    Polyurethane Polyurethane Non-woven Regarding the inks, one conductive silver ink (Table 6) and three types of dielectric inks have been employed ( Table 7). The inks influence noticeably the final capacitance at a thickness level as well as at a relative permittivity level. Heat sealed polyurethanes (Table 8) have been employed as well in some sensors as substitutes of the dielectric inks.
Once the different sensors were built, they have been characterized electrically by means of the determination of their C TxRx , C RxGND , and C TxGND capacitances. Lastly, the sensitivity of the different sensors has been determined using a graphical user interface software from Microchip, AUREA.

Materials
Different cotton, polyester and mixed fabrics with different fabric densities, yarn diameter, and weave have been used. Tables 4 and 5 show the main characteristics of the fabrics used.
The conductive ink (Table 6) used is silver IPC-603X from INKRON (Kutojantie, Espoo, Finland). Three dielectric inks (Table 7) with different flexibility and stretchability characteristics have been used. The inks are 127-48D from CREATIVE (Lamesa, Texas, USA), DI-7542 from EMS (Delaware, Ohio, USA) and IPD-670 from INKRON. Two polyurethane films (Table 8) have been used. The films are EU94DS from DELSTARD and UAF445 from ADHESIVE FILMS.

Sensor Development
Manufacturing technology used was based on serigraphic technology of thick film. The screen-printing process consists of forcing pastes of different characteristics over a substrate through some screens using squeegees. Openings in the screen define the pattern that will be printed on the substrate by serigraphy. The final thickness of the pastes can be adjusted by varying the thickness of the screens. When screen-printing technology is used, it is necessary to manufacture frames with screen mesh for the design.
The screen for the conductors was a 230 mesh polyester material (PET 1500 90/230-48 from Sefar) and the screen for dielectric layer was a 76 mesh polyester material (PET 1500 30/76-120 PW from Sefar). Afterwards, to transfer the pattern to screen mesh, a UV film Dirasol 132 from Fujifilm was used. The final screen thickness was 74 µm for the screen for conductors and 217 µm for the screen for the dielectric. The patterns were transferred to the screen by using a UV light source unit IC-5000 from BCB.
Printing was carried out by using E2XL from EKRA screen-printer with a shore 75 • hardness squeegee, 60 • squeegee angle, 1 mm snap-off, 3.5 bar force, and 100 mm/s. After the deposition of the inks, these were cured in an air oven FED-115 from BINDER at 130 • C for 15 min in order to use the same curing characteristics for all inks. In the case of DI-7542 from EMS a UV oven Ncure-Lab/Static 120 from EneMaq was used with 0.5 J/cm 2 .
The polyurethanes have been heat sealed on the fabrics with a DCH-100 heatpress from Microtec at 130 • C for 60 s.

Measurements
The capacitance between electrodes was measured at 100 kHz with a KEYSIGHT U1733C LCR meter.
The relative permittivity (ε r ) measures were carried out with a Hewlett Packard 4263A LCR meter. The following measurement accessories were used: Hewlett Packard 16089B Kelvin Clips Leads and a Yokogama-Hewlett Packard 16451A Dielectric Test Adaptor. The LCR meter was configured to measure a tension level of 1V, with an average of 64 samples and a low read rate (Level = 1 V, Avg = 64, Meas Time= Low). The measurement mode was Cp and D (parallel capacity and loss tangent). The capacity measurements were taken at three different parts of the fabric with a 4-frequency scan (0.1 kHz, 1 kHz, 10 kHz, and 100 kHz). The ε r value was obtained directly from the Cp value.
The measurements of the thickness of the fabric were taken at 4 different points using a Mitutoyo CP calibre CD-6" with a 10 µm resolution.
Macroscopic images were taken with a LEICA MZ APO stereomicroscope.

Results and Discussion
In the 3DS-1 design, the dielectric layer between the Rx electrodes and the Tx electrode must mandatorily be a screen-printed layer of dielectric ink, since this layer contains the connection traces. Implementing these traces with vias in a layer made with textile is practically impossible. Thus, in this design, the substrate textile can be used as the dielectric layer between Tx layer and GND layer (Figure 17a). The textile provides its physical and electrical characteristics, or it can only be utilized as the base substrate (Figure 17b) where it only contributes with its physical characteristics. Figure 18 shows the final sensor, the appearance being the same in both of the implemented structures. The polyurethanes have been heat sealed on the fabrics with a DCH-100 heatpress from Microtec at 130° C for 60 s.

Measurements
The capacitance between electrodes was measured at 100 kHz with a KEYSIGHT U1733C LCR meter.
The relative permittivity (εr) measures were carried out with a Hewlett Packard 4263A LCR meter. The following measurement accessories were used: Hewlett Packard 16089B Kelvin Clips Leads and a Yokogama-Hewlett Packard 16451A Dielectric Test Adaptor. The LCR meter was configured to measure a tension level of 1V, with an average of 64 samples and a low read rate (Level = 1 V, Avg = 64, Meas Time= Low). The measurement mode was Cp and D (parallel capacity and loss tangent). The capacity measurements were taken at three different parts of the fabric with a 4frequency scan (0.1 kHz, 1 kHz, 10 kHz, and 100 kHz). The εr value was obtained directly from the Cp value.
The measurements of the thickness of the fabric were taken at 4 different points using a Mitutoyo CP calibre CD-6" with a 10 μm resolution.
Macroscopic images were taken with a LEICA MZ APO stereomicroscope.

Results and Discussion
In the 3DS-1 design, the dielectric layer between the Rx electrodes and the Tx electrode must mandatorily be a screen-printed layer of dielectric ink, since this layer contains the connection traces. Implementing these traces with vias in a layer made with textile is practically impossible. Thus, in this design, the substrate textile can be used as the dielectric layer between Tx layer and GND layer (Figure 17a). The textile provides its physical and electrical characteristics, or it can only be utilized as the base substrate (Figure 17b) where it only contributes with its physical characteristics. Figure 18 shows the final sensor, the appearance being the same in both of the implemented structures.  As aforementioned, the CRxTx, CTxG, and CRxG capacitances must be of a low value and within a very concrete range. Once the dimensions of the areas of these capacitors are fixed, there are only two parameters to be modified to achieve a low capacitance. These parameters are the layer thickness and the relative permittivity. For this reason, the three selected dielectrics have been characterized looking for the one with the lowest relative permittivity. Regarding the layer thickness, a large thickness would be desirable. To increment the thickness, the physical parameters of the screen-print must be varied. The mesh must be low in the screen to augment the quantity of ink to be printed and several layers must be screen-printed successively to increase the thickness. Table 9 shows the relative permittivities of the three available inks, where CREATIVE 12-48D presents the lowest relative permittivity. The right column shows the minimum thickness recommended by Microchip to achieve an appropriate value of CRxTx. These large values of thicknesses cannot be obtained using screen-print techniques. Table 9. Relative permittivity of the dielectric inks εr @100 kHz. A first prototype with type A substrate was implemented to characterize the sensor. Two sensors were implemented: a sensor named 3DS-1a-TA and another one named 3DS-1b-TA. To be able to estimate the capacitances using theoretical calculus, the value of the relative permittivity of the textiles is needed. Table 10 shows the determined values. The right column shows the minimum thickness recommended by Microchip to achieve an appropriate value of CRxTx, considering the permittivity obtained from the textiles. Type A textile is located well below the limit; Type C textile is located below as well, and the rest are on the limit or exceed it widely.  As aforementioned, the C RxTx , C TxG , and C RxG capacitances must be of a low value and within a very concrete range. Once the dimensions of the areas of these capacitors are fixed, there are only two parameters to be modified to achieve a low capacitance. These parameters are the layer thickness and the relative permittivity. For this reason, the three selected dielectrics have been characterized looking for the one with the lowest relative permittivity. Regarding the layer thickness, a large thickness would be desirable. To increment the thickness, the physical parameters of the screen-print must be varied. The mesh must be low in the screen to augment the quantity of ink to be printed and several layers must be screen-printed successively to increase the thickness. Table 9 shows the relative permittivities of the three available inks, where CREATIVE 12-48D presents the lowest relative permittivity. The right column shows the minimum thickness recommended by Microchip to achieve an appropriate value of C RxTx . These large values of thicknesses cannot be obtained using screen-print techniques. Table 9. Relative permittivity of the dielectric inks ε r @100 kHz.

Dielectric
Relative Permittivity t > ε r /5 (µm) A first prototype with type A substrate was implemented to characterize the sensor. Two sensors were implemented: a sensor named 3DS-1a-TA and another one named 3DS-1b-TA. To be able to estimate the capacitances using theoretical calculus, the value of the relative permittivity of the textiles is needed. Table 10 shows the determined values. The right column shows the minimum thickness recommended by Microchip to achieve an appropriate value of C RxTx , considering the permittivity obtained from the textiles. Type A textile is located well below the limit; Type C textile is located below as well, and the rest are on the limit or exceed it widely.  Table 11 shows the different values obtained from 3DS-1a-TA for C TxRx , C RxGND , and C TxGND . The average dielectric thickness obtained between the Rx and Tx layers is 20 µm. Type A textile has an average thickness of 110 µm. The average thickness between Rx and GND is 135 µm.  Table 12 shows the different values obtained from 3DS-1b-TA for C TxRx , C RxGND , and C TxGND . The average dielectric thickness obtained between Rx and Tx layers is 40 µm. The average thickness between Tx layer and GND is 60 µm. The average thickness between Rx layer and GND is 100 µm. The capacitances obtained with the 3DS-1a-TA sensor as well as with 3DS-1b-TA, widely exceed the margins recommended by Microchip. This is fundamentally due to the fact that the reached thicknesses with the screen-printed dielectric layers do not achieve the minimum values recommended by Microchip. These high capacitance values affect the input and output signals from/to the controller. Figure 19 shows the waveforms from the 3DS-1a-TA sensor. The transmission signal (Figure 19a) presents a deformation due to a capacitance of C TxGND > 1 nF. If a buffer operational amplifier is coupled between the Tx pin and the Tx electrode, a regenerated signal is obtained (Figure 19b). Figure 19c shows the receiving Rx signal with direct connection between the Tx pin and the Tx electrode. Figure 19d shows the receiving Rx signal with an AO between the Tx pin and the Tx electrode. The results obtained with the 3DS-1b-TA sensor are very similar to those obtained with the PCB reference sensor (Figure 12). These capacitance values out of the recommended range have forced the use of the 3DS-2 sensor. This sensor allows us to modify the thickness as well as the relative permittivity of the dielectric Rx-Tx since the textile substrate is used as dielectric ( Figure 16). When studying the response using the same textile substrate (Type A) but in 3DS-2 configuration, a significant reduction of the capacitance values is observed (Table 13), although the values are still out of the range recommended by Microchip. This high value out of the range comes from the yet insufficient dielectric thickness, since the type A textile has a thickness of 110 μm. For this reason, textiles with a noticeably higher thicknesses have been employed, between 380 and 1300 μm. In this design, the connection lines of the Rx electrodes with the connector are located on the same layer as the Rx electrodes. Thus, the lines influence the total capacitance. In the case of the North electrode, the connection line is out of the sensitive area and has least influence, getting a lower capacitance with respect to the rest.  These capacitance values out of the recommended range have forced the use of the 3DS-2 sensor. This sensor allows us to modify the thickness as well as the relative permittivity of the dielectric Rx-Tx since the textile substrate is used as dielectric ( Figure 16). When studying the response using the same textile substrate (Type A) but in 3DS-2 configuration, a significant reduction of the capacitance values is observed (Table 13), although the values are still out of the range recommended by Microchip. This high value out of the range comes from the yet insufficient dielectric thickness, since the type A textile has a thickness of 110 µm. For this reason, textiles with a noticeably higher thicknesses have been employed, between 380 and 1300 µm. In this design, the connection lines of the Rx electrodes with the connector are located on the same layer as the Rx electrodes. Thus, the lines influence the total capacitance. In the case of the North electrode, the connection line is out of the sensitive area and has least influence, getting a lower capacitance with respect to the rest. The rest of the textiles were fabrics with large thicknesses and presented much roughness. This led to problems with the printing of the conductive and dielectric inks. A couple of samples are presented in Figure 20. On the left, a printing on a type D textile can be observed, and, on the right, a printing on a type E textile. In both textiles, the conductive ink is not uniformly distributed on the textile, leading to very high resistance values, or even infinite, i.e., open circuits or with no electrical continuity. This problem appears on Type B, C, D, E, and F textiles, but on the contrary, in Type G, H, and I, it is possible to screen print the conductive silver ink with no errors of continuity.  This problem can be solved modifying the surface of the textile to soften the roughness. The modification consists of the printing of a dielectric layer on the textile or the heat sealing of a polyurethane film. In the case of the printing of a dielectric layer, the three available dielectrics have been used on each one of the textiles. Later, a conductive silver ink layer was printed to check if there is any improvement in the result of the printing. Figure 21 shows the result of the printing of the different dielectric inks: (a) Type D textile with dielectric Creative 127-48D and a layer of silver ink, (b) Type D textile with dielectric EMS DI-7542 and a layer of silver ink, (c) Type D textile with dielectric Inkron IPD-670 and a layer of silver ink, (d) Type E textile with dielectric Creative 127-48D and a layer of silver ink, (e) Type E textile with dielectric EMS DI-7542 and a layer of silver ink and (f) Type D textile with dielectric IPD-670 and a layer of silver ink. This problem can be solved modifying the surface of the textile to soften the roughness. The modification consists of the printing of a dielectric layer on the textile or the heat sealing of a polyurethane film. In the case of the printing of a dielectric layer, the three available dielectrics have been used on each one of the textiles. Later, a conductive silver ink layer was printed to check if there is any improvement in the result of the printing. Figure 21 shows the result of the printing of the different dielectric inks: (a) Type D textile with dielectric Creative 127-48D and a layer of silver ink, (b) Type D textile with dielectric EMS DI-7542 and a layer of silver ink, (c) Type D textile with dielectric Inkron IPD-670 and a layer of silver ink, (d) Type E textile with dielectric Creative 127-48D and a layer of silver ink, (e) Type E textile with dielectric EMS DI-7542 and a layer of silver ink and (f) Type D textile with dielectric IPD-670 and a layer of silver ink. The problem of the lack of continuity in the conductive tracks could not be solved with the dielectrics. This led to the use of heat sealable polyurethanes on the different textiles. Specifically, EU94DS from Delstar Inc., and UAF-445 from Adhesive Films. These polyurethanes (Table 8) can help to improve the capacities of the sensor providing a larger layer thickness (80 μm in the case of EU94DS and 120 μm the case of UAF-445). In addition, they can improve the relative permittivity. Table 14 shows the permittivities measured for each one of the used polyurethanes. The printing of the conductive ink on the textiles covered with polyurethanes is perfect. With the employment of the polyurethanes, the structure of the 3DS-2 sensor changes depending on the textiles. For Type G, H, and I textiles, the structure of Figure 22a (named 3SD_2a) has been employed and for Type B, C, D, E, and F textiles, the structure of Figure 22b (named 3SD_2b) has been employed. The problem of the lack of continuity in the conductive tracks could not be solved with the dielectrics. This led to the use of heat sealable polyurethanes on the different textiles. Specifically, EU94DS from Delstar Inc., and UAF-445 from Adhesive Films. These polyurethanes (Table 8) can help to improve the capacities of the sensor providing a larger layer thickness (80 µm in the case of EU94DS and 120 µm the case of UAF-445). In addition, they can improve the relative permittivity. Table 14 shows the permittivities measured for each one of the used polyurethanes. The printing of the conductive ink on the textiles covered with polyurethanes is perfect. With the employment of the polyurethanes, the structure of the 3DS-2 sensor changes depending on the textiles. For Type G, H, and I textiles, the structure of Figure 22a (named 3SD_2a) has been employed and for Type B, C, D, E, and F textiles, the structure of Figure 22b Table 15 shows the values of the capacitances associated to each one of the types of developed sensors. A remarkable decrease of the capacitance values can be noticed. This is due to the increase of the thickness of the substrates and the values of the relative permittivities. The sensor that most met expectations was 3DS-2a-TI with very low values of CTxRx and CRxGND capacitances, far below the recommended limit by Microchip. Regarding the value of the CTxGND capacitance, it is situated above the recommended value (<1nF). It can be solved inserting an op-amp between the Tx pin and the Tx electrode as recommended by Microchip.      Table 15 shows the values of the capacitances associated to each one of the types of developed sensors. A remarkable decrease of the capacitance values can be noticed. This is due to the increase of the thickness of the substrates and the values of the relative permittivities. The sensor that most met expectations was 3DS-2a-TI with very low values of CTxRx and CRxGND capacitances, far below the recommended limit by Microchip. Regarding the value of the CTxGND capacitance, it is situated above the recommended value (<1nF). It can be solved inserting an op-amp between the Tx pin and the Tx electrode as recommended by Microchip.   Table 15 shows the values of the capacitances associated to each one of the types of developed sensors. A remarkable decrease of the capacitance values can be noticed. This is due to the increase of the thickness of the substrates and the values of the relative permittivities. The sensor that most met expectations was 3DS-2a-TI with very low values of C TxRx and C RxGND capacitances, far below the recommended limit by Microchip. Regarding the value of the C TxGND capacitance, it is situated above the recommended value (<1 nF). It can be solved inserting an op-amp between the Tx pin and the Tx electrode as recommended by Microchip. Figure 24 shows the waveform of the 3DS-2a-TI sensor. Figure 24a shows the Tx signal with a buffer OA inserted between the Tx pin and the Tx electrode due to the value of CTxGND, greater than 1 nF. Figure 24b shows the signal in one of the Rx electrodes when there is no object modifying the electrical field. The Tx signal was the original with no buffer. Figure 24c shows the same Rx signal after inserting the buffer in the path of the Tx signal. Microchip provides the AUREA graphical user interface that allows for characterizing the sensors. The sensitivity is, maybe, the most important parameter of these sensors. Microchip [43] provides an "artificial hand", a 40 × 40 × 70 mm styrofoam (εr ≈ 1) cube covered with an adhesive copper sheet. This block must be connected to ground to simulate the conditions of the human body.
To determine the sensitivity, the block is placed at different distances from the surface of the sensor using blocks of styrofoam of different thicknesses (1, 2, … cm) ( Figure 25). AUREA allows us to read the obtained data. Figure 26 shows a representation of the data. It can be observed that as the associated capacitances increase, the signal deviation decreases for the same distance to the surface. The 3DS-2a-TI sensor presents the best sensitivity. The sensitivity is even better than the one of the PCB's sensor from Microchip, due to the low value of their associated capacitances. Microchip provides the AUREA graphical user interface that allows for characterizing the sensors. The sensitivity is, maybe, the most important parameter of these sensors. Microchip [43] provides an "artificial hand", a 40 × 40 × 70 mm styrofoam (ε r ≈ 1) cube covered with an adhesive copper sheet. This block must be connected to ground to simulate the conditions of the human body. To determine the sensitivity, the block is placed at different distances from the surface of the sensor using blocks of styrofoam of different thicknesses (1, 2, . . . cm) (Figure 25). AUREA allows us to read the obtained data. Figure 26 shows a representation of the data. It can be observed that as the associated capacitances increase, the signal deviation decreases for the same distance to the surface. The 3DS-2a-TI sensor presents the best sensitivity. The sensitivity is even better than the one of the PCB's sensor from Microchip, due to the low value of their associated capacitances.  It would be convenient to be able to determine the value of the associated sensor capacities prior to their manufacturing. Thus, only those textiles that achieve capacities below 30 pF would be utilized in the sensors.  It would be convenient to be able to determine the value of the associated sensor capacities prior to their manufacturing. Thus, only those textiles that achieve capacities below 30 pF would be utilized in the sensors. It would be convenient to be able to determine the value of the associated sensor capacities prior to their manufacturing. Thus, only those textiles that achieve capacities below 30 pF would be utilized in the sensors.
Two theoretical values have been calculated, the nominal theoretical value (C n ) calculated with Equation (1) and the edge effect capacitance value (C edge ). This latter value considers the effect of the field lines around the edges of the capacitor and can be calculated according to Equations (2) and (3).
where C is the value of the capacitance in pF, L is the length in cm, w is the width in cm, t is the thickness in cm, ε r is the relative permittivity, and ε 0 is the vacuum permittivity (8.85 × 10 −12 F/m). A comparative study has been made with the North electrode. This electrode is the one with less influence of the capacitances associated to the Rx conduction lines to the connector. Table 16 shows the values of the theoretical capacities and the real C TxRx value measured with an LCR meter. In general, the theoretical results are validated by the real ones. The C edge value is very close to the real one in all the cases. Thus, it is possible, knowing the relative permittivity of each textile and its thickness, to deduct the value of its associated capacitances prior to the sensor manufacturing. The following figures show the frames of a video with some gestures that the controller can recognize: Figure 27. Approach detection.; Figure 28. Flick north to south.; Figure 29. Flick west to east.; and Figure 30. Airwheel. The videos are available as Supplementary Materials (Videos S1 and S2).         A Bluetooth portable system has been developed to be used with mobile devices (Figure 31). The system is configured with the AUREA application and the configuration parameters are saved in the portable device. To check the functioning, it is connected to an Android device as wireless mouse as shown in Figure 32. A Bluetooth portable system has been developed to be used with mobile devices (Figure 31). The system is configured with the AUREA application and the configuration parameters are saved in the portable device. To check the functioning, it is connected to an Android device as wireless mouse as shown in Figure 32. A Bluetooth portable system has been developed to be used with mobile devices (Figure 31). The system is configured with the AUREA application and the configuration parameters are saved in the portable device. To check the functioning, it is connected to an Android device as wireless mouse as shown in Figure 32.

Conclusions
The aim of this research was to obtain a wearable textile 3D gesture recognition sensor based on the Microchip 3D GestIC ®® sensor design. The fundamental parameters of that sensor design are the capacitance between transmission and reception electrodes. The value of this capacitance depends fundamentally on the thickness of the substrate and its relative permittivity. Therefore, the key to the design is to look for the best textile substrate materials. But the best textile material is not always the best material to be used with screen-printed technology. For this reason, a large proportion of the research has consisted of trying different fabrics. Fabrics are composed of different materials and, therefore, with different thickness and relative permittivity but also with different behaviour regarding silver and dielectrics inks. A flexible textile 3D sensor was developed after studying all these parameters.
That sensor was characterized in relation to its electrical parameters, namely its capacitances between electrodes, its thickness, and its sensitivity to the presence of a hand. The characteristics

Conclusions
The aim of this research was to obtain a wearable textile 3D gesture recognition sensor based on the Microchip 3D GestIC ®® sensor design. The fundamental parameters of that sensor design are the capacitance between transmission and reception electrodes. The value of this capacitance depends fundamentally on the thickness of the substrate and its relative permittivity. Therefore, the key to the design is to look for the best textile substrate materials. But the best textile material is not always the best material to be used with screen-printed technology. For this reason, a large proportion of the research has consisted of trying different fabrics. Fabrics are composed of different materials and, therefore, with different thickness and relative permittivity but also with different behaviour regarding silver and dielectrics inks. A flexible textile 3D sensor was developed after studying all these parameters.
That sensor was characterized in relation to its electrical parameters, namely its capacitances between electrodes, its thickness, and its sensitivity to the presence of a hand. The characteristics obtained were better than Microchip's PCB design and with the further advantage of its flexibility. The Microchip driver (MGC3XXX) incorporates a calibration procedure that can be programmed, for example, on stand-by periods. This allows the driver to adjust to the base signal according to the bending angle. This feature maximizes the advantage of using a flexible sensor.
We are currently working on the study of the variation of two important features of textiles, such as softness or permeability. The use of soft polyurethane films guarantees softness. Regarding the permeability, the addition of different layers reduces it. We are currently exploring different solutions such as microperforations and hole patterns. Lastly, we are preparing some experiments to determine the washing capacity of the sensor, due to the importance of washing in the textile area.
In conclusion, a 3D gesture sensor based on E-field change technologies has been developed to be used with textile substrates using a low cost and common textile industry printing technique: screen-printing. The system works on both flat and curved surfaces, which allows it to be used in several areas: clothes, automobile industry, healthcare, etc.