Motion Discrimination from Multi-channel EMG Signals Using Crosstalk Information between Electrodes

Abstract

This paper proposes a novel motion discrimination method for human interface equipments. Using crosstalk electromyography (EMG), which is defined as difference between EMG signals acquired from two different muscles, motions can be discriminated with fewer electrodes than the standard bipolar electrode configuration. In this method, even a single channel of the crosstalk EMG signal acquired from a pair of muscles can provide information from multiple muscles. In order to achieve accurate discrimination, frequency features of the crosstalk EMG signals extracted by a set of filter bank and stochastic characteristics of the features are effectively utilized using a probabilistic neural network (PNN). Experimental results show that eight motions of the forearm can be discriminated with the proposed method using the crosstalk EMG acquired from three electrodes. Comparing with a traditional method, which is based on bipolar electrode configurations, it is found that the proposed method can achieve considerably higher discrimination ability using only half of electrodes.