Elsevier

Pattern Recognition

Volume 110, February 2021, 107648
Pattern Recognition

Multimodal subspace support vector data description

https://doi.org/10.1016/j.patcog.2020.107648Get rights and content
Under a Creative Commons license
open access

Highlights

  • A novel method for transforming the multimodal data into a common feature space is proposed.

  • The shared subspace optimized for one-class classification yields better results than traditional concatenation of multimodal data for one-class classification.

  • Different regularization strategies along with linear and non-linear formulation provides more freedom of choice for optimizing a model according to specific evaluation metric.

Abstract

In this paper, we propose a novel method for projecting data from multiple modalities to a new subspace optimized for one-class classification. The proposed method iteratively transforms the data from the original feature space of each modality to a new common feature space along with finding a joint compact description of data coming from all the modalities. For data in each modality, we define a separate transformation to map the data from the corresponding feature space to the new optimized subspace by exploiting the available information from the class of interest only. We also propose different regularization strategies for the proposed method and provide both linear and non-linear formulations. The proposed Multimodal Subspace Support Vector Data Description outperforms all the competing methods using data from a single modality or fusing data from all modalities in four out of five datasets.

Keywords

Feature transformation
Multimodal data
One-class classification
Support vector data description
Subspace learning

Cited by (0)

Fahad Sohrab is a PhD student in Unit of Computing Sciences, Tampere University, Finland. He received his MS degree in Electronics Engineering from Sabanci University, Istanbul, Turkey in 2016. His research interests include machine learning, pattern recognition, and anomaly detection.

Jenni Raitoharjuis a Senior Research Scientist in Programme for Environmental Information at Finnish Environment Institute. Finland. She received her PhD in Information Technology in Tampere University of Technology, Finland in 2017. Her current projects deal with machine learning and pattern recognition in applications such as biomonitoring and autonomous systems.

Alexandros Iosifidis received his PhD degree in Informatics from the Aristotle University of Thessaloniki in 2014. He is an Associate Professor of Machine Learning at Aarhus University, Denmark. His research interests include statistical machine learning and artificial neural networks with applications in Computer Vision and time-series analysis problems.

Moncef Gabbouj received his MS and PhD degrees in electrical engineering from Purdue University, in 1986 and 1989, respectively. Dr. Gabbouj is Professor of Signal Processing at the Department of Computing Sciences, Tampere University, Finland. His research interests include Big Data analytics, multimedia analysis, artificial intelligence, machine learning, and pattern recognition.