Automated Speaker Recognition Using Compressed Temporal- Spectral Dynamics Information of Password Spectrograms
Amitava Das · 2008
Prevalent speaker recognition methods use only spectralenvelope based features such as MFCC, ignoring the rich speaker identity information contained in the temporalspectral dynamics of the entire speech signal. We propose a new feature for speaker recognition called compressed spectral dynamics (CSD) which effectively captures such spectral dynamics and the inherent speaker identity. The discriminative power of CSD allows the classification part to remain simple. The proposed method, a simple nearest neighbor classifier using CSD, delivers performance competitive to conventional MFCC+DTW based textdependent speaker recognition methods at significantly reduced complexity.