Large margin filtering for signal sequence labeling

TL;DR

This paper introduces a joint learning approach for filtering and classification in signal sequence labeling, optimizing filter parameters to improve accuracy in noisy signal data.

Contribution

It proposes a novel large margin filtering method that jointly learns the filter and classifier, including optimal cutoff frequency and phase, for enhanced signal sequence labeling.

Findings

Effective on toy and real BCI datasets

Outperforms naive filtering methods

Learns optimal filter parameters for noise reduction

Abstract

Signal Sequence Labeling consists in predicting a sequence of labels given an observed sequence of samples. A naive way is to filter the signal in order to reduce the noise and to apply a classification algorithm on the filtered samples. We propose in this paper to jointly learn the filter with the classifier leading to a large margin filtering for classification. This method allows to learn the optimal cutoff frequency and phase of the filter that may be different from zero. Two methods are proposed and tested on a toy dataset and on a real life BCI dataset from BCI Competition III.