波形变换域语音波形的帧间差分编码

Intra-frame Differential Coding for Speech Waveform in Time Domain

本文提出一种新的语音波形的帧间差分编码算法，文中对帧间差分编码所遇到的帧匹配、过渡帧处理和量化器设计等问题进行了讨论，文末给出了汉语普通话几个常见单元音量化信噪比的测试结果。实验表明：采用四比特帧间差分编码，当信号采样频率为８ｋＨｚ时，其短时量化信噪比的平均值超过２９ｄＢ，利用归一化新波形的对称性、半周期取零技术和基音重复，预测误差用４比特量化时，可在数码率２ｋ～４ｋｂｐｓ情况下得到优质重建语音。

We propose a new intra-frame differential coding algorithm for speech waveform intime domain,This coding algorithm treats the waveforms in pre-proceeding as two-dimensionsignals.X-axis signal is the sequence number of samples in the same frame and Y-axis is thesequence number of frames,which performs the matching differential coding for the two sequentframes.Problems caused by intra-frame differential coding,such as frame matching,transitionframe proceeding and quantizer design,are discussed in detail.SNR test results for the somevowels of Mandarin Chinese is presented.The experiment shows that the short-time quantizationSNR of 4-bits intra-frame differentizal coding surpases 2 9 dB at 8kHz sampling rate. It isshown that there is a possibiliry to systhesis excellent speech with bit rate of 2 ～ 4kbps.