Automatic spoken language identification by digital signal processing methods. Tatar and Russian languages

Abstract

The paper studies the problem of language identification for audio files. For solving the problem, we use methods of digital signal processing only (without analysis of phonemes distinctive for language). A special attention is drawn to the form of signal in an area close to the position of a stop consonant. The evaluation is performed on a set of two languages; this includes speech records taken from TV programs. It is provided that solely one of the two languages (either Tatar or Russian) is used in each of files. Experimental evidence demonstrates the feasibility of the proposed techniques.