Realization of Audio Discrimination System Based on VQ
In the process of speech control, there is a problem that may cause the incorrect operation of the system, independent non-speech signals will easily be taken for speech signals. This paper will suggest a way to differentiate the non-speech signals and speech signals by vector quantization (VQ), and recognize the specific types of speech signals in the further for facilitate post-processing. First, the system extracts the Mel Frequency Cepstrum Coefficient (mfcc) from the collected audio waveform sequences to get the feature vector sequence of audio. Next we use the lbg algorithm of vector quantization to obtain the template of voice signal and non-voice signal respectively, and make a distortion measure between the feature vector sequence of audio and each template, then select a template type with smallest quantization error for audio type. By this way, we discriminate and determinate the type of speech signals. In this paper, we processed the system simulation on matlab, the result shows that the recognizable rate of this system is over 87％, performed the function of differentiate the nonspeech signals and speech signals, achieved better purpose.
VQ Distortion Measure mfcc lbg algorithm matlab
Chen Baoyuan Wang Xiaoliang HaoYanli You Xilan
The higher educational key laboratory for Measuring & Control Technology and Instrumentations of Heilongjiang Province,Harbin University of Science and Technology. Harbin 150040 China
国际会议
厦门
英文
105-108
2010-05-22（万方平台首次上网日期，不代表论文的发表时间）