Filter bank learning for signal classification

Maxime Sangnier (1) , Jérôme Gauthier (1) , A. Rakotomamonjy (2)

1 LS2D - Laboratoire Sciences des Données et de la Décision
2 LITIS - Laboratoire d'Informatique, du Traitement de l'Information et des Systèmes

This paper addresses the problem of feature extraction for signal classification. It proposes to build features by designing a data-driven filter bank and by pooling the time-frequency representation to provide time-invariant features. For this purpose, our work tackles the problem of jointly learning the filters of a filter bank with a support vector machine. It is shown that, in a restrictive case (but consistent to prevent overfitting), the problem boils down to a multiple kernel learning instance with infinitely many kernels. To solve such a problem, we build upon existing methods and propose an active constraint algorithm able to handle a non-convex combination of an infinite number of kernels. Numerical experiments on both a brain-computer interface dataset and a scene classification problem prove empirically the appeal of our method.

Keywords: Discriminative feature-extraction; Scattering; Recognition; Design; Signal classification; Optimization; Support vector machine; Kernel learning; Filter bank; SVM; signal processing; machine learning; artificial intelligence; classification

Domains: Signal and Image Processing; Machine Learning [stat.ML]; Applications [stat.AP]; Metric Geometry [math.MG]

Maxime Sangnier, Jérôme Gauthier, A. Rakotomamonjy. Filter bank learning for signal classification. Signal Processing, 2015, 113, pp.124 - 137. ⟨10.1016/j.sigpro.2014.12.028⟩. ⟨cea-01865050⟩