基于小波变换和矢量量化的语音压缩编码方案

Speech compression scheme based on wavelet transform and vector quantization

对于一种新的语音压缩编码方法——基于小波变换和矢量量化的语音压缩编码方案进行了开发和研究。该方法先对语音进行小波变换和小波包变换，然后使用矢量量化来压缩数据．我们还研究了使用摘编码方法进一步降低比特率的算法，实验结果表明对于采样频率为８ｋＨｚ，每样本量化为８ｂｉｔ的６４ｋｂｉｔ／ｓ语音信号压缩到６ｋｂｉｔ／ｓ～８ｋｂｉｔ／ｓ时，具有良好的语音质量．同时该算法还具有低延迟的优点，其算法延迟只有８ｍｓ。

A new coding method of speech compression which is based on Wavelet Transform and Vector Quantization(VQ), is developed and studied in this paper. Firstly the speech is processed with Wavelet Transform or Wavelet PacketTransform, then VQ is used to compress the coefficients of Wavelet Transform. The entropy coding is also used todecrease the bit rate. The experimental results demonstrate that the speech signal sampled by 8 kHz sampling rateand 8 bit quantization, so that 64 kbit/s bit rate, can be compressed to 6 kbit/s ~ 8 kbit/s, and still have good speechquality. The low--delay is another advantage for this method, which is only 8 ms.