模式匹配和过零率检测的音频差错掩盖

Audio Error Concealment Based on Pattern Matching and Zero-Crossing Detection

针对用传统波形替代方法恢复音频波形变化趋势不准确的问题,将模式匹配和短时过零率检测结合应用于音频的差错掩盖研究中。在对丢帧后的音频信号数据进行模式匹配的掩盖基础上,将从模式匹配模块输出的音频信号数据和参考信息一同送入短时过零率检测模块,进一步对丢失数据进行掩盖,最终将从短时过零率模块输出的参考帧作为丢失帧的替代,从而得到音频信号的重构。与传统方法相比,在音频信号的重构效果上,获得了更好的恢复音频。实验证明,该算法对音频差错的掩盖更加精确,信噪比值平均提高了0.3～0.5 dB,主观语音质量评估值平均提高了0.3。

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.