为了降低移动网络丢包、抖动失真等现象对音频质量的影响,提出一种基于帧间相关性的导谱频率系数的差错隐藏算法。该方法根据近邻成功接收若干好帧的导谱频率(ImmittanceSpectralFrequencies,ISF)距离预测与当前坏帧的相关性,再基于相关性进行导谱频率系数的自适应坏帧恢复。解决了现有算法采用单一的坏帧恢复技术,在面临多样化的音频信号输入时,坏帧恢复效果不佳的问题。在1%~10%的误码条件下的客观实验结果表明,该算法谱失真明显优于AVS-P10(AVS移动音频编码标准)的差错隐藏算法,性能平均得到42%的提升。主观实验也同时表明了该算法的优势。该方法可扩展应用到音频编码器的自适应码本及固定码本等参数的差错隐藏。
Error concealment for hybrid Algebraic Code Excited Linear Prediction(ACELP)and transform codecs is proposed to reduce the effect of the packet loss and network jitter during the audio transmission.When the frame is not successfully received, its Immittance Spectral Frequencies(ISF)coefficients are reproduced using previous good frames’ISF coefficients. The number and perceptual weight of the previous usable good frames are decided by their correlation to the current frame in ISF distance. The original error concealment methods usually use fixed number of previous good frames to recover the missing frame. When facing with versatile audio inputs, the proposed adaptive algorithm is more effective compared with the original fixed number methods. The objective results show that the proposed algorithm has 42% reduction of spectral distortion for AVS-P10 codec at error frame rate from 1% to 10%. The subject result also validates the advantage of this paper. The algorithm can be used in the error concealment of other parameters like adaptive and fixed codebook.