Speaker Dependent Coefficients for Speaker Recognition

This work aims at speaker recognition based upon a new set of features. Feature extraction is a crucial phase of the recognition process and a proper feature set dramatically influences the speaker recognition. Many well-known features are not suitable for the speaker recognition as those merge the specifics of the individual voices. Therefore, we need features accentuating the individual differences of our voices to be able to recognise speakers reliably. This work introduces new, speaker dependent features called Speaker Dependent Frequency Cepstrum Coefficients (SDFCC), created for the speaker recognition purposes only. Experimental results show their performance in comparison to the well-known features. According to the test results, the SDFCC are, for the speaker recognition, very useful and promising.