An Accurate and Inter-Operatable Fuzzy-Based System Using Genetic and Canonical Correlation Analysis Methods in Internet-of-Brain Things

The brain computer interface is defined as the way of acquiring the brain signals that analyse and translates them into commands that are relayed to intelligent devices for carrying out various actions. Through number of BCI mechanism and approaches have been proposed by various scientists to empower the individuals for directly controlling their objects via their thoughts. However, the actual implementation and realization of this method faces number of challenging with low accuracy and less interoperability. In addition, the pre-processing signals and feature extraction process is further time consuming and less accurate. In order to overcome the mentioned issues, this paper proposes an accurate and highly inter-operable system using genetic fuzzy system along. The predictive model and analysis can be further improved using canonical correlation analysis. The proposed framework is validated and demonstrated using brain typing system analysis. The results are computed against accuracy, latency and interoperability of the signals received from brain with less SNR along with traditional method. The proposed mechanism shows approximately 87% improvement as compare to existing approaches during the simulation over various performance metrics.

sector of communication including several domains such as home and city automation, industrial manufacturing, healthcare and transportation [1], [2]. Individuals have wide variety of opportunities to a control and interact with a wide range of objects through various means of applications running on wearable devices, smartphones, gestures or brain-computer interface [3], [4]. The Brain Computer Interface (BCI) is considered as one of the latest paradigms in order to provide an interaction among individuals and intelligent devices using a direct pathway among brain and external devices. In addition, the recent research tells that the BCI have witnessed the translation of brain thinking into physical actions such as IoT enabled appliances which brings the BCI a major aiding technology in human interaction systems [5], [6]. Though BCI considered as a cognitive interactive facility having several benefits and advantages of controlling the human brains and interact the brain signals through intelligent devices. However, it may further lead to several interactive challenges. The signals of brain that can be measured through various techniques and technologies such as magnetoencephalography (MEG), Electroencephalogram (EEG), all of these are susceptible to less trust worthy with influenced environmental factors [7], [8]. The Internet-of-Brain Things (IoBT) is considered as one of latest paradigm for collecting the signals and behavior of each individual through intelligent devices using direct pathway among devices and brain. The brain signals having low signalto-noise ratio lack the temporal and spatial resolution on deep insights of brain activities. As a result, the recognition system and interaction among brain and intelligent devices can achieve upto 80% of accuracy that is insufficient for practical demonstrations. In addition, the data pre-processing, selection of features and extraction both in frequency and time domain consumes a lot of time that is further dependent on human expertise. Furthermore, the collection of information such as imaging data needs an accurate predictive model for gathering the information by linking the brain activity and behavior.
The depicted Figure 1 shows the activities of brains that can be easily identified and recognized by various medical techniques such as Magnetic Resonance Imaging (MRI), electroencephalogram (EEG) and Functional near-infrared spec- troscopy (FNIRS). The following techniques can be further used to recognize the various activities of brain that can be further recorded and stored via various online servers and store them in database. The recorded and stored activities of brain using various technologies now can be handled and managed through recent modern techniques. The Internet-of-Things where each and every device acts smartly can be easily integrated to any other application such as healthcare, society and education etc. In this Figure 1, the recorded and stored activities of brain can be easily managed and handled via various smart technologies and machines to efficiently recognize the human brain. The collected brain signals through EEG or MEG are further broadcasted to the online server via internet. The online server further uses a fuzzy system and canonical correlation analysis for accurate analysis of the brain signals. The analysed results can be further interpreted to actuate the functions in various applications such as transportation control, smart cities, assistive robot control and so on [9]. Fuzzy systems have been considered the more reliable solution to ensure an accurate and correct predictive model for gathering and controlling the activities of human brains using intelligent devices. Number of fuzzy systems have been proposed by various researchers and scientists such as simple fuzzy systems having higher interoperability are not usually accurate while the complex fuzzy system having higher interoperability are not accuracy [10], [11], [12]. Therefore, multi-objective optimization algorithms have been to used to find and accurate and highly interoperable system.

A. Motivation
The contribution of fuzzy systems is diverse depending on various areas of applications including collection of brain signals through intelligent devices by ensuring a direct pathway among brain signals and external devices. BCI is defined as one of the trending phenomena for collecting the brain activities for controlling and interacting with a wide range of objects through various means of applications running on wearable devices, smartphones, gestures, or brain-computer interface. However, the smooth integration among neurons and external devices sometimes leads to uncertainty with less accuracy and analysis of signals due to poor generalization, invariants, and bad SNR. Genetic fuzzy systems have been considered as one of the efficient computational frameworks for ensuring the accurate analysis of brain signals via external devices. In addition, neuroimaging can be considered along with fuzzy systems for further improving the prediction ratio of brain activities through intelligent devices. Though number of fuzzy systems and fuzzy logic have been proposed by various scientists/researchers, such as Nojima et al. [13] have proposed multiple fuzzy partitions using multi objective genetics machine learning mechanism using search ability. In addition, Zhang et al. [14], Hsu et al. [15] and Rahim et al. [16] have proposed a brain computer interface by enabling cognitive interactivity using selective attention method. However, the higher interoperability along with more accuracy system is yet to be explored in IoT environment. In addition, the brain emotional learning mechanism is provided using fuzzy control through chaos synchronization. Furthermore, the neuroimaging predictions using multi-output can be analysed using canonical correlation and reduced rank regression analysis.

B. Contribution
The proposed framework aim is to interpret, analyse and predict the brain signals and activities based upon genetic fuzzy systems and canonical correlation analysis method. In this paper, we have integrated the two different mechanisms for ensuring the more accurate and predictive model for analysing the brain activities through intelligent devices. The proposed framework is capable of modelling a high inter-operable and high accurate signal capturing mechanism for interpreting the information. The main contribution of the proposed framework is highlighted as follows.
• A genetic fuzzy logic system is proposed to interpret the individuals brain activity for enabling the IoBT interaction using intelligent devices.
• The genetic fuzzy system is further integrated with canonical correlation analysis method that is specifically used in neuroimaging for better prediction of model. The canonical analysis discovered the most distinguishable feature that is further broadcasted to the system for better analysis.
• The proposed mechanism is demonstrated using brain typing system prototype for validating the practicality and efficiency of the proposed approach. The remaining organization of the paper is structured as follows. The related section of discussed introduction with various approach and techniques proposed by various researchers is illustrated in section II. An accurate and highly interoperable interaction mechanism among with fuzzy system and canonical correlation analysis is discussed in section III. Further, the performance of proposed mechanism is validated and illustrated in section IV. Finally, section V concludes the paper and directs the future scope of the work.

II. RELATED WORK
This section deliberates the number of techniques and approaches proposed by various researchers and scientists for ensuring an accurate and reliable communication among devices using various fuzzy and AI based methods in IoBT approaches as presented in Table I. Zhang et al. [14] have proposed a unified deep learning framework for ensuring an accurate interpretation of individual brain signals and IoT devices. The authors have designed a reinforcement learning mechanism using selective attention by proposing a modified lon short-term memory. they have analysed the efficiency by conducting a real-world demonstration and experimentation in comparison of traditional approaches. Coogan and He [17] have created a diverse application by providing the end-user autonomy during the distribution of task at real time. The authors have demonstrated the validity of proposed mechanism using virtual reality unity-based environment and commercial devices through BCI2000. Qin et al. [18] have proposed an adaptive reinforcement learning algorithm based upon Bayesian optimization method for balancing the optimality and speed of convergence. The authors have evaluated the proposed prototype using tensor flow system by showing the outperformance against various traditional approaches. The authors claimed that the proposed approaches reduced the latency and violation up to 88.9% and 91.6% respectively.
Jingtao et al. [19] have analysed the present status of information centric network and then realized the security issues of information centric network in internet of brain things. The authors have proposed a secure architecture mitigating from privacy and various threats. They have proposed a coaching method based on popularity based and authentication method based on interval. The proposed approach is verified and equipped with simulation-based platform based on various security threats and mechanisms. Abdulghani et al. [20] have proposed a brain waves deep learning recognition method based on colors and shapes for merging the concept of brain computer interface and internet of things. The proposed mechanism is verified by showing the results against accuracy and recognition of reliability of designs. Yang et al. [21] have proposed a brain like productive system using federated learning-based prediction system. the proposed mechanism is based upon three mechanisms globally optimized reservation of resources, graph-based mining and federated learning prediction. The proposed framework is predicted and analysed against various measuring metrics such as quality of service and accuracy. Rao el al. [22] have compared various methods for dealing or predicting the nuisance variables for neuroimaging the data. The authors have adjusted the imaging information using regression method with an addition of predicators with separate kernel. The authors have evaluated the structural and f MRI information by discussing the modelling mechanisms. Hasanzadeh and Kasaei [23] have proposed a fuzzy system based on Bayesian and ML for stimulating the brain signals. The authors have validated the demonstrated results for medical image segmentation by showing the strengths and limitations.
Though consumer electronics is termed as one of oldest and significant research where scientists have been working on it for years for providing a comfortable and ease of life to their consumers. However, the security and privacy of consumer electronics is still at its early stage. In addition, very few of the authors have discussed about security concerns while combining and operation large number of smart devices in the network. In order to ensure a secure and trusted consumer electronic system, we have proposed a trusted and secure architecture by proposing the TOPSIS and blockchain technology for sensing, evaluating and analysing the communicated device in the network. Figure 2 represents the overall framework of the proposed mechanism that includes two key components individual and Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply. environment. The depicted Figure 2 generated the input data by recognizing various activities of brain using latest medical approaches and methods. The recorded input data is now integrated with two separate methods known as fuzzy systems and canonical correlation analysis. The genetic fuzzy system is used to recognize the each and every activity of human nerve to take an accurate and correct decision. In addition, the prediction of brain activities can be identified by integrating The canonical correlation analysis method with genetic fuzzy system. The environment component includes fuzzy system and canonical correlation analysis. The fuzzy system and canonical model determine the state transition and reward model while individual model represents the agent to collect brain signals. The detailed explanation of both the components is defined as follows.

A. Fuzzy System
The system model of proposed framework is discussed in Figure 2 having input data, fuzzy systems and CAA approach. The fuzzy system is used to derive the detection value of each sector from the given set of signals. The signals that are defined as inputs are 1) sector area and 2) reinforcement ratio. In case the detection value of the fuzzy system is greater than a significant value then the analysis is considered as accurate otherwise inaccurate. In addition, the detection value is high in case of high reinforcement ratio and less ratio. Figure 2 depicts the fuzzy systems inputs and output variables. The detailed explanation of each and every component is further illustrated in below text.

B. Genetic Fuzzy System
The genetic algorithms in fuzzy systems are the one that use the operations searched in natural genetics to guide the trek using search space. The genetic algorithms are proven to ensure the reliable searching capabilities in complex spaces. The genetic fuzzy system allows the exploration of various characteristics of same object over different parameters. The characteristic is measured by forming an image to further complete the segmentation process. The proposed algorithm is based on pixels classifications by training the fuzzy classifier for each cell through evolution process. The classification rule of fuzzy system is based upon knowledge base system as follows: L1P1is¬m1and P2is¬m2and P Dis¬m3theninputϵtoc. (1) where, P1, P2 and PD are the pixel intensities in images and M1, M2 and M3 are the membership functions by defining the class name.
Further, the output of the fuzzy system is determined using final classifier by merging all the classifier results. The final classifier is further prepared by mixing the gaussian functions with the trained data. Two probability distribution functions are designed for the classifier C. the one PD is for firing the classifier degree training the samples of C and another is for sampling the remaining classes as follows: P D c (P) = P(P = p|class = c) P D ¬ c(P) = P(P = p|class ̸ = c) The genetic algorithm is further improved by using the If-then rules of knowledge-base to get the brain signals in accurate manner. If we have n number of received brain signals B R P = (br p1 , · · · B R pn ) where P = 1, 2 · · · m from n different devices D where B R pi derives the attribute value of p th pattern of i th device. The IF-then rule can be further defined as: L q i f m1is M q1 andm n is M qn thenclassc q withcw q . ( where L q is the q t h fuzzy rule label and M q1 is the antecedent fuzzy set (i = 1, · · · n), C q is class label and C W q is the weight of class label. The C q and C W q of each fuzzy rule are specified from various input brain signals along with their A q . The confidence and accuracy of each class label is computed as: Authorized licensed use limited to the terms of the applicable license agreement with IEEE. Restrictions apply.
where µAq(P p ) is the compatibility grade with antecedent A q of P p that is further computed as: µA q (P p ) = µA q1 (P p1 ) . . . µA qn (P pn ) (4) Finally, the C W q of each fuzzy set R q used the following specifications:

C. Canonical Correlation Analysis
In order to further maximize the accuracy with reduced latency, CCA is used to improve the correlation between P ai and Q b i where P belongs to real feature matrix of n X p and Q belongs to score matrix. The maximized correlation ration of CCA is defined as: where the presented formula illustrates the variant of reduced rank regression and CCA coefficients a j , b j are defined as singular value decomposition as: where MCCA is the diagonal matrix of square roots of corresponding k eigenvalues. HCCA are the first k eigenvectors of normalized RCCA. The CCA estimates the coefficients by maximizing the correlations between P and Q and discarding the input and output variance. The CCA reduces the latency of analysis of prediction model by decomposing the process of RCCA. The working of the entire proposed mechanism can be easily understood using an Algorithms 1 and 2 as explained below.

Algorithm 1 Secure Algorithm
Prerequisite: 10% of devices are altered from legitimate to malicious by the intruders upon increasing the network size. All the communicating devices are legitimate upon establishing the network Input Value: (1) A network N having N = n 1 , n 2 , n 3 · · · n n number of IoT devices Output: Device is legitimate or altered Given: An accurate analysis model using Genetic Fuzzy model and Canonical Correlation Analysis Step 1: Establish the networking environment. all nodes I d = I = 1, 2, · · · D p=1 to N Compute accuracy of each device using estimated Genetic Fuzzy () Model (Device is ideal) Maintain a genetic fuzzy () network of legitimate devices and determine their accuracy Block/deny further communication Step 2: Each ideal device is predicted using CCA Algorithm 2 Genetic Fuzzy() and CCA() Model Step 1: The output of the fuzzy system is determined using final classifier by merging all the classifier results: P D c (P) = P(P = p|class = c) P D ¬ c(P) = P(P = p|class ̸ = c) µA q (P p ) = µA q1 (P p1 ) . . . µA qn (P pn ) (10) Step 2: The confidence and accuracy of each class label is computed as: Finally, the C W q of each fuzzy set R q used the following specifications: Step 3: The maximized correlation ration of CCA is defined as: IV. PERFORMANCE ANALYSIS This section illustrates the designing of synthesized experiment to analyse the efficiency and validity of proposed mechanism. The experiments are initially recorded and compared against traditional baselines to analyse the overall performance of the system. In addition, the proposed framework is investigated against latency and accuracy parameters that are defined as two crucial metrics while analysing the signals via intelligent devices.

A. Experimental Details
The type of network that we have considered here for validating the proposed phenomenon is heterogeneous where any type of device may gather, and analyse the information in an environment. The EEG is collected suing portable and commercialized Epoc and headset containing 12 channels with the sampling rate of 128 Hz. The experiment is carried out over 4 subjects (2 male and 2 female) aged 24-29 corresponding to the hints appears on their screen. In addition, the brain signals are collected if subjects are imaging some actions such as left arrow, right arrow, up or down arrow etc. the imaginary things are labelled along with their brain actions using Table I. The experiments contained total 150 750 samples with 23 150 samples. The classification results are further analysed over recall, precision, accuracy and latency of the brain signals detection via intelligent devices. In order to further analyse the efficiency of proposed framework, some traditional mechanisms are also considered.

B. Baseline Methods
The baseline models that are considered for verifying and validating the proposed mechanism is discussed in this section. Abdulghani et al. [20] have proposed a brain waves deep learning recognition method based on colors and shapes for merging the concept of brain computer interface and internet of things. The proposed mechanism is verified by showing the results against accuracy and recognition of reliability of designs. Yang et al. [21] have proposed a brain like productive system using federated learning-based prediction system. the proposed mechanism is based upon three mechanisms globally optimized reservation of resources, graph-based mining and federated learning prediction. The proposed framework is predicted and analysed against various measuring metrics such as quality of service and accuracy. The authors have validated the demonstrated results for medical image segmentation by showing the strengths and limitations. Table II represents the overall comparison of proposed framework with the traditional mechanism. In addition, the key parameters of traditional approaches are listed as SVM and KNN as c=1, n=150 and k = 2. The observations in Table II shows that the proposed mechanism outperforms all the traditional approaches by showing more accuracy and less latency using genetic fuzzy system and canonical correlation analysis method. The accuracy, ROC curves, precision and latency results of proposed framework are reported in Figure 4, 5, 6 and 7. Furthermore, the simulation set up of proposed framework is further mentioned in Table II Figure 3 represent the accuracy of proposed mechanism as compare to [20] and [21] the accuracy of proposed framework is improved by further integrating the genetic fuzzy system with CCA as it may provide more reliable and high interoperability as compare to traditional mechanisms. Figure 4 depicts the ROC curves of various classes while collecting the brain signals through genetic method. The extra noise and in-built noises are removed automatically to further improve the signal ratio through intelligent devices. Figure 5 represents the precision of the proposed framework that determines and validates the selection of genetic algorithm over any other fuzzy systems. The genetic algorithm improves the reliability and precision rate of brain signals that may further improves the accuracy of the data via intelligent devices.   Finally figure 6 illustrates the latency curve of the proposed framework. The integration of genetic method with CCA reduces the noise ad SNR ratio by eliminating the distorting while gathering the brain signals from the individuals.

C. Results and Discussion
The proposed mechanism is further analysed over reliability of the network that behaves ideal during unethical behaviour in the network. The reliability measures the failure tolerance  of the system that can behave ideally and provide efficient and significant results in the network. It is defined as the number of times the system behaves ideally during altered nature or environment of the system.

Reliabilit y(t)
where I (t) is defined as total number of requests processed by ideal devices over a period of time t, T (t) determines the total number of requests ethical or non ethical processes by the devices (either ideal or non ideal devices). The reliability of the system determines the failure tolerance of the system during un-ideal conditions or situations in the environment. The proposed system as depicted in Figure 7 reliability is much better as compare to other two existing approaches because of canonical correlation analysis mechanism that predicts the behaviour or idleness of the system during transmission of information in the network. The integration of genetic fuzzy system and canonical correlation analysis makes the system much more reliable and efficient. Figure 7 shows a reliability comparison of the proposed mechanism over the traditional method. The indirect trust evaluation and reinforcement learning measures and examines the malicious behavior of the devices. The proposed mechanism outperforms the traditional scheme due to a continuous computation and surveillance of malicious devices, blocking them immediately to prevent further communications in the network.
The adaptability of the system is determined by analysing their communication behaviour. The systems adaptability is analysed based upon the number of requests processed by the system successfully. The percentage of the requests fulfilled by the system upon increasing the traffic (data) in the network measures the system performance.
where Adaptabilit y(t) determines the number of requests R(t) fulfilled by the system among total number of requests T (t) processed by the devices. The graph of adaptability can be easily determined from the Figure 8 in comparison of existing approaches. The adaptability of proposed approach is outperformed as compare to existing mechanism because of their genetic fuzzy behaviour that behaves according to the dynamic nature of the system.

V. CONCLUSION
We proposed a genetic fuzzy system along with canonical correlation analysis (CAA) method for improving the accuracy with reduced latency while collecting the brain signals of the individuals via intelligent devices. The internet of brain things is considered as one of latest paradigm for collecting the signals and behavior of each individual through intelligent devices. The proposed mechanism significantly outperformed as compare to existing mechanism and is analysed and validated against several traditional methods of collecting the signals. The proposed framework is validated against accuracy, latency, precision and ROC while analysis and predicting the information because of genetic behavior of fuzzy system. The reliability and latency is further improved by integrating the genetic mechanism through canonical correlation analysis. The proposed mechanism showed approximately 87% improvement as compare to existing mechanism. The proposed framework is further experimented over real time data sets along with various case studies to check the validity and feasibility in the future directions of this paper.