Biometric Authentication based on EMG Signals of Speech

Electromyography (EMG) of speech is the recording of muscle action potentials in the neck region during the speech. This paper proposed a first of its kind method of a biometric system using EMG of a speech signal. Signals were denoised by using empirical mode decomposition (EMD). All the noise was removed leaving us with the regions of interest. Next, feature extraction was performed using a combination of time and frequency-based features. The classification was then performed using different classifiers like Support Vector Machines (SVM), K-nearest neighbor (KNN), Ensemble classifiers, and Decision Tree (DT). The results showed that the maximum accuracy of 95.3% was obtained by using the Quadratic SVM classifier for all 10 classes.