Collaborative Learning for Incremental Classification of EMG Signals

In this work, we address the problem of online learning for electromyogram (EM G) classification. The main challenge of EMG is its streaming nature, which implies that not all data are available at once. As such, offline and batch methods such as deep learning are generally not appropriate for such small data. To address this problem, we propose an incre-mental learning technique whereby the algorithm enhances its knowledge about the data progressively, as it receives new data. Moreover, we propose to employ multiple learners (in-stead of a single learner); each learning and focusing on a specific class. This federated learning strategy in incremental learning shows superior performance over both existing incre-mental learning and deep learning methods in our simulation studies on small data.