Informatica

New text independent speaker identification method is presented. Phase spectrum of all-pole linear prediction (LP) model is used to derive the speech features. The features are represented by pairs of numbers that are calculated from group delay extremums of LP model spectrum. The first component of the pair is an argument of maximum of group delay of all pole LP model spectrum and the second is an estimation of spectrum bandwidth at the point of spectrum extremum. A similarity metric that uses group delay features is introduced. The metric is adapted for text independent speaker identification with general assumption that test speech channel may contain multiple speakers. It is demonstrated that automatic speaker recognition system with proposed features and similarity metric outperforms systems based on Gaussian mixture model with Mel frequency cepstral coefficients, formants, antiformants and pitch features.

The isolated word speech recognition system based on dynamic time warping (DTW) has been developed. Speaker adaptation is performed using speaker recognition techniques. Vector quantization is used to create reference templates for speaker recognition. Linear predictive coding (LPC) parameters are used as features for recognition. Performance is evaluated using 12 words of Lithuanian language pronounced ten times by ten speakers.

The problem of speaker identification is investigated. Basic segments – pseudostationary intervals of voiced sounds are used for identification. The identification is carried out, comparing average distances between an investigative and comparatives. Coefficients of the linear prediction model (LPC) of a vocal tract, cepstral coefficients and LPC coefficients of an excitation signal are used for identification as features. Three speaker identification methods are presented. Experimental investigation of their performance is discussed.

The use of vector quantization for speaker identification is investigated. This method differs from the known methods in that the number of centroids is not doubled but increases by 1 at every step. This enables us to obtain identification results at any number of centroids. This method is compared experimentally with the method (Lipeika and Lipeikienė, 1993a, 1993b), where feature vectors of investigative and comparative speakers are compared directly.