Voice Based Person Identification Using c-Mean and c-Medoid Clustering Techniques

Digital systems interfaced networks predominantly govern present day society. Authentication of an individual is of greater importance for security and protection of a robust network-based world. Biometric recognition is one of the interests for researchers to cater this need. Person identification based on voice recognition using mean clustering and medoid clustering approaches are implemented and analyzed here. The pre-emphasized speech signal acquired by microphone is subjected to silence removal and endpoint detection. The time series speech waveform thus obtained is subjected to MFCC and pitch computation. Three classes of c-means: Hard c-means, fuzzy c-means, and rough c- means algorithms are implemented and compared with three classes of medoid algorithms: hard c-medoid, fuzzy c-medoids, and rough c-medoid. The fuzzy c-medoid algorithm outperforms the other voice-recognition systems in terms of recognition accuracy.