Sparse representation for time-series classification

This chapter studies the problem of time-series classification and presents an overview of recent developments in the area of feature extraction and information fusion. In particular, a recently proposed feature extraction algorithm, namely symbolic dynamic filtering (SDF), is reviewed. The SDF algorithm generates low-dimensional feature vectors using proba- bilistic finite state automata that are well-suited for discriminative tasks. The chapter also presents the recent developments in the area of sparse- representation-based algorithms for multimodal classification. This in- cludes the joint sparse representation that enforces collaboration across all the modalities as well as the tree-structured sparsity that provides a exible framework for fusion of modalities at multiple granularities. Fur- thermore, unsupervised and supervised dictionary learning algorithms are reviewed. The performance of the algorithms are evaluated on a set of field data that consist of passive infrared and seismic sensors.