Characterizing EMG Signals using Aggregated CENSUS Transform

This research work aims to propose a systematic procedure for characterizing Electromyogram (EMG) signals. This algorithm efficiently extracts the inherent patterns present in the physiological signals, by exploiting the self-similarity of the signal. A finite set of aggregated processes are created from the raw EMG signal. Now, CENSUS transform values are calculated for each of these aggregated processes. Finally, the CENSUS transform values form the feature vector and this can further be utilized for characterization and classification of the EMG signals. For assessing the utility of the proposed feature extraction technique, extensive experiments are conducted on two publically available datasets using k-NN classifier. It is evident from the results that our method achieves higher classification accuracy than many of the state-of-the-art techniques.