Abstract:

The traditional waveform substitution techniques for recovering missing audio segments can not accurately reconstruct the trend of the audio waveform. The theory and the algorithm of pattern matching and shorttime zero-crossing are applied to the research of the audio error concealment method. When frames of the audio are lost, the audio signal data is sent into the module of pattern matching. After the procedure of pattern matching, the audio signals with data and reference information are processed in the module of short-time zero-crossing. So it is synthesized as the reference frame to substitute the lost frame in order to reconstruct the original audio signal. Compared with the traditional audio signal waveform reconstruction, this proposed method is better for audio frame reconstruction. The experimental results show that the average SNR（Signd-to-Noise） of the reconstructed audio is increased by 0.3—0.5 dB and the average PESQ(Perceptual Estimation of Speech Quality） of the reconstructed audio is increased by 0.3.

Key words: audio, error concealment, pattern matching, short-time zero-crossing