An approach to content-independent feature extraction for Chinese-Korean spoken language identification

A new classification feature extraction method for Chinese-Korean spoken language identification was proposed in this paper. Firstly, speech signal was divided into frame serial and the number of frames was counted. Furthermore, the ratio between short-time zero-crossing rate and short-time energy, i.e. short-time-frequency-energy-ratio (STFER), was computed, and the mean STFER per frame was treated as the classification feature to implement Chinese-Korean spoken language identification. Finally, the classification threshold was determined using information gain. Experimental results show that the proposed method is simpler than MFCC feature parameters and has better ability to identify spoken language with lower complexity, can be adopted in preprocessing procedure of language recognition.