LOW-COMPLEXITY AUTOMATIC SPEAKER RECOGNITION IN THE COMPRESSED GSM AMR DOMAIN (WedAmOR2)

Author(s) : Matteo Petracca (Politecnico di Torino, Italy) Antonio Servetti (Politecnico di Torino, Italy) Juan Carlos De Martin (IEIIT - CNR, Italy)

Abstract : This paper presents an experimental implementation of a low-complexity speaker recognition algorithm working in the compressed speech domain. The goal is to perform speaker modeling and identification without the need of decoding the speech bitstream to extract speaker dependent features, thus saving important system resources, for instance, in mobile battery powered DSP devices. Bitstream values of the GSM AMR speech coder parameters are studied to identify sufficient statistic that enables fair recognition after few seconds of speech. Euclidean distance measures on elementary statistic variables such as coefficient of variation and skewness obtain recognition accuracy of 100% after 20 seconds of active speech for a database of 14 speakers recorded in a normal room environment.

Menu