Speech recognition using facial sEMG

Abstract

This paper presents a study of speech recognition based on electromyographic biosignals captured from the articulatory muscles in the face using surface electrodes. This paper compares the speech recognition system for spoken English and Malay words by a group of Malay native speakers. Feature extraction was done in both temporal and time-frequency domains. Temporal features used are integrated EMG (IEMG), mean absolute value (MAV), root mean square (RMS), variance (VAR), standard deviation (SD), and simple square integral (SSI) where time-frequency domain features were obtained using discrete wavelet transform. For classification, random forest classifier and ANNs multilayer perceptron both gave the overall best performance on using temporal features and time-frequency features respectively. The result of the classification shows that the Malay language is can be used in sEMG speech recognition.