A study on the validity of speaker identification using sound color marker

Abstract

The development of IT technology, such as smart devices and pin-tecs, is leading to convenient and highly secure biometric authentication. In biometric authentication, voice authentication is a very efficient biometric authentication method in terms of convenience, security, and economy. The voice is expressed through the complex energy distribution and change in the sense of personality and words. In this paper, we use the Sound Color Marker calculated from the time series average spectrum of voice signals to verify the validity of speaker identification. In experiments with adults 10 speech, extraction time for reference sound source is not less than 60 seconds, extraction time for the sound source to be discriminated if more than 10 seconds, the speaker identification rate using the Sound Color Marker was very high. Also, even if we do not know what it says, we found that the longer the extraction time for the reference sound source and the extraction time for the sound source to discriminate, the higher the speaker identification rate. These results show that context-independent type speaker identification is possible only by comparing the sound color markers, which are simply expressed by seven parameters. © International Research Publication House.