基于信号规整和稀疏变换的语音与音频分层编码方法

The Layered Coding of Speech and Audio Signals Based on Signal Warp and Sparse Transform

基于语音和音频信号的固有周期性特征,本文构建了一种适合语音和音频信号的统一分析/合成模型,并分别在24kbps和32kbps码率下,实现了对宽带语音和音频信号的高质量分层编码.首先,本文将具有时变周期的输入信号规整为具有固定周期的信号,并对规整后的周期信号构建规整矩阵;其次,对规整矩阵的行和列分别进行调制叠接变换(MLT)和离散余弦变换(DCT),完成规整矩阵的稀疏化;最后,利用分带量化和矢量哈夫曼编码完成稀疏矩阵元素的量化和编码.主客观测试结果表明,本文所提方法的语音、音频及其混合信号的编码质量均优于同等速率下的ITU-T G.722.1和AMR-WB编码器.

Based on the periodic characteristics of speech and audio,a layered coding method by using uniform analysis and synthesis model is proposed in this paper. The constructed coder can perform equally well on speech and audio at the bit rates of 24 kbps and 32 kbps. First,the input signal which has time-varying period is warped into a constant period signal.Second,a sparse representation of the warped signal is achieved by applying the MLT and DCT on the warped matrix derived from the warped signal. Finally,the sub-band quantization and Huffman coding are applied on the transform coefficients. Both the objective PESQ / PEAQ results and the subjective A / B listening tests showthat the proposed coder outperforms the ITU-T G. 722. 1 and AMR-WB codec.