Speaker Identification System Using Lpc -Application On Berber Language
F. Z. Chelali et al. · Zenodo (CERN European Organization for Nuclear Research) · 2015
The aim of our article is to design a speaker identification system applied to the Berber language, based on linear predictive coding parameterization and maximum likelihood classification. Our objective is to establish a system for speaker identification in dependent and independent mode using continuous speech with duration of 2 or 3 seconds. To ensure that task, we have decided to employ some features based on the Linear Predictive Coding (LPC). The used classifier is based on the maximum likelihood principle. Furthermore, a speech dataset from Berber language and using emphatic syllables was recorded in laboratory and analysed off-line. The experimental evaluations were performed using the first and second derivatives of the descriptors, and the fundamental frequency (i.e. pitch). Results show that the proposed features and approach are interesting in speaker identification.