Speaker recognition using lower formants

Speaker recognition is the area of speech technology, which attempts to identify the speaker. There are two distinct approaches in this domain. One is based on blind DSP processing where we deal with the signal regardless of its content. The other is phonetic based where individual phonemic units are identified, processed and matched on the basis of certain features. Our research is related to the second approach and we have tried to identify high level phonemic features like style etc. from the processing of lower formants. We have tried to process these formants on transition boundaries and attempted to demonstrate that each person has a specific pattern of rise and fall in the transition region. Our basic hypothesis is: the transition of formants at the consonant-vowel boundaries is controlled by the brain and it contains a specific signature of the speaker. Experiments have shown that our method has a success rate of 90%. Moreover, the system gives desirable results, even in a noisy environment due to the choice of lower formants