Voice activity detection algorithm based on Mel frequency cepstrum coefficient(MFCC) similarity

To improve the accuracy of Voice Activity Detection (VAD) under noisy Environment, a voice activity detection algorithm based on Mel Frequency Cepstrum Coefficient (MFCC) similarity is proposed. First, the MFCC from each frame of the voice signals is extracted. Then, the first ten frames are taken as the background noises. Finally, by calculating the MFCC correlation coefficient distance under the above noisy condition, the voice-activity parameters are detected from the MFCC similarity curves. The experiment results show that the proposed algorithm is effective under both white noise and pink noise conditions and at low signal to noise ratio.