Development of corpora for person recognition using humming, singing and speech

There is very good research potential for developing speech processing applications in Indian languages (such as speech and speaker recognition). In this paper, we describe the development of speech corpora in Hindi (an Indian language) with three different voice biometric patterns, i.e., humming, singing and speech. Difference in the voice biometric patterns, for the application as automatic person recognition system is also described in very brief. Our experimental results indicate that the person-dependent characteristics are distributed in different manner for each voice biometric signal. We also observed that person recognition system performs relatively better in the case of humming as compared to the singing and speech biometric signal.