Title :
Perceptual features in speaker identification
Author :
Segarceanu, Svetlana ; Zaharia, Tiberius ; Radoi, Constantin
Author_Institution :
Dept. of Appl. Electron. & Inf. Eng., Politeh. Univ. of Bucharest, Bucharest, Romania
Abstract :
Although different spectral representations carry similar information, in practice various features differ in their performances. In this paper we will examine two perceptual approaches, dwelling upon the perceptual analysis based on the bark scale. Their ability to separate classes is assessed with the help of the within-class and between-class matrices.
Keywords :
matrix algebra; speaker recognition; between-class matrices; speaker identification; spectral representations; within-class matrices; Algorithm design and analysis; Autocorrelation; Band pass filters; Cepstral analysis; Feature extraction; Filter bank; Fourier transforms; Frequency; Humans; Speech analysis; linear prediction; perceptual analysis of speech; similarity/dissimilarity; speaker identification; vector quantization;
Conference_Titel :
Communications (COMM), 2010 8th International Conference on
Conference_Location :
Bucharest
Print_ISBN :
978-1-4244-6360-2
DOI :
10.1109/ICCOMM.2010.5509069
