Investigation of the relation between amount of VoIP speech data and performance in speaker identification task over VoIP networks

VoIP is a promising technology that has been expected to be the future of voice telecommunications. Thus, with the increasingly widespread use of internet-enabled devices, the interest for speaker recognition systems over VoIP networks will be increased day by day. In this paper, we intend to analyze and investigate the relation between the amount of VoIP speech data, optimal speaker's model size and performance in text-independent speaker identification over VoIP networks. All experiments performed in this study were conducted under a state-of-art speaker identification system, using Mel frequency cepstral coefficients (MFCC) as acoustic features and Gaussian mixture models for speaker modeling.