Automatic Language Identification in speech streams

The problem of Automatic Language Identification (ALI) is a very interesting area of research up today. The objective of this work is to propose a new approach based on acoustic parameters in order to identify six languages. These parameters must be adapted to the dynamic nature of speech in order to improve the automatic identification of languages. The experiences are led with six languages (German, English, Arabic, Spanish, French and Italian) on a corpus including recordings extracted from some conversations of various movies and some missions of the television and of radio. The results found show that the use of these new parameters can provide higher identification rate than those obtained with the best acoustic parameters previously used.