A Real Time Language and Dialect the Vq Error of the Acoustic Fea

This paper proposed an application for the language and dialect recognition in spontaneous speech on the real time. The language will be identified based on the VQ (Vector Quantization) error and the fundamental frequency. And the dialect recognized based on the VQ error and the sample different in speaking rate. The reported application system in this paper performed on the seven languages in Asia: Chinese, Japanese, Korean, Thai, Mongolian, Uigur and Khazak; And the dialect performed on the four dialects of Mongolian, such as Kharha of the Mongolia, Chahar of the Inner Mongolian, Oirat of the Xin Jiang in China, and Kalmyk of Kalmykia in Russia. Results from a number of experiments showed that the misidentification rate for language is less than 3% in less than 3 seconds utterance.