Novel text independent speaker recognition using LPC based formants

Abstract

Speaker recognition is the process of identifying a speaker by his/her speech samples. By extracting the speaker-specific features from the speech samples, the recognition task can be done. The formant estimation of speech sample of specific speaker is important for feature extraction in speaker recognition, because the formants are unique and reflect the vocal tract information of a speaker. In the proposed work Linear Predictive Coding (LPC) based formant estimation method is employed as it is more efficient and accurate. The formant frequencies of each speaker are evaluated and stored in the database and test speaker speech sample is also processed in the same way to find the formants and finally they are compared with the formants stored in the database to perform the speaker verification and recognition.