Time Series Clustering Algorithm for Two-Modes Cyclic Biosignals

In this study, an automatic algorithm which computes ameanwaveis introduced. Themeanwaveis produced by averaging all cycles of a cyclic signal, sample by sample. With that information, the signal’s morphology is captured and the similarity among its cycles is measured. A k-means clustering procedure is used to distinguish different modes in a cyclic signal, using the distance metric computed with themeanwaveinformation. The algorithm produced is signal-independent, and therefore can be applied to any cyclic signal with no major changes in the fundamental frequency. To test the effectiveness of the proposed method, we’ve acquired several biosignals in context tasks performed by the subjects with two distinct modes in each. The algorithm successfully separates the two modes with 99.3% of efficiency. The fact that this approach doesn’t require any prior information and its preliminary good performance makes it a powerful tool for biosignals analysis and classification.