Automatic Language Identification using spectrum characteristics and Bessel funetions

Automatic language identification is a significant context of speech processing . There have been numerous studies on automatic language identification context, in majority of which the extracted characteristic from speech signal ,PLP or MFCC were two factors. In this study,a new language identification system is introduced in which characteristic exctraction will be based on Bessel Fourier transform factor and a new WRBF naurotic network as well as RBF Network are used in its category.Results achieved from the new system were compared with results achieved through known methods of PLP&MFCC characteristic extraction as well as MLP neurotic network. Results of assays performed on OGI database and pair by pair comparison of speeches depict a significant accuracy of language detection of RBF & WRBF network rather than MLP network. Also,with respect to the fact that the accuracy of PLP&MFCC methods are so close to Fourier-Bessel transform characteristic Extraction, this type of characteristic Extraction can be introduced as a powerful method in this context.