EMG based voice recognition

Besides its clinical applications, various researchers have shown that EMG can be utilised in areas such as computer human interface and in developing intelligent prosthetic devices. This paper presents results from a preliminary study. The work describes the outcomes in using artificial neural network (ANN) to recognise and classify human speech based on EMG. The EMG signals were acquired from three articulatory facial muscles. Three subjects were selected and participated in the experiments. Preliminarily five English vowels were used as recognition variables. The root mean square (R S) of EMG signals were estimated and used as set of features to feed the ANN. The findings indicate that such a system may have a capacity to recognise and classify speech signals with an accuracy of up to 88%.