基于DSP的EVRC声码器实现与优化

【作者】 李强；

【导师】 张捷；

【作者基本信息】 西北工业大学 ， 信号与信息处理， 2004， 硕士

【摘要】 为了满足第三代移动通信系统对高质量语音和大容量系统的需要，在CDMA系统中普遍采用了变速率语音编码技术，根据话音的信号能量和背景噪声动态地决定编码速率，这种方法既保证了话音的质量，又降低了平均编码速率，从而增加了系统容量。 TIA于1997年正式推出IS-127标准，它是IS-95的可选语音服务标准(Service Option 3)和第三代移动通信系统CDMA2000的语音编解码标准，此标准的核心算法是增强型可变速率语音编码器(Enhanced Variable Rate Codec，简称EVRC)。EVRC声码器是可变速率的，最大编码速率为8kbps，在话音质量上接近于QCELP-13k的语音编码器，且具有更好的抗干扰能力。在目前的CDMA系统中，综合语音质量和编码速率，EVRC是最佳的语音编码器。 在IS-95和CDMA2000的基站中，EVRC声码器都得到了广泛的应用。由于我们开发的声码器主要用于基站，所以对性能的要求较高，选用了TI公司的最先进的DSP芯片TMS320C64xx。本文围绕定点EVRC声码器的软件优化，做了大量的工作，在充分理解TMS320C64xx CPU结构的基础上，针对语音信号处理中大量出现的循环运算进行了深度优化，大大提高了CPU的利用率以及CPU功能单元的并行程度，同时，我们还用TIA/EIA/IS-718的测试向量对定点EVRC声码器进行了严格比特对准测试，提高了声码器的鲁棒性。 最后，将我们开发的声码器与Delphi公司开发的声码器基准程序(Benchmark)作了比较，结果表明，我们的声码器的性能要明显高于Delphi公司的，这使得EVRC声码器具有较大的应用价值。更多还原

【Abstract】 To meet the needs of the third mobile communication system for higher speechquality and greater system capacity, the CDMA systems employ variable rate speechcoding technologies, which select the encode rate dynamically according to the signalenergy and background noise. This method insures the speech quality as well as reduces,the average encode rate, so the system capacity is increased.In 1997, TIA has released the standard of IS-127, it is Service option 3 of IS-95 and the voice encoder/decoder standard of the third mobile communication, the core algorithm of this standard is Enhanced Variable Rate Codec, the shortened form is EVRC. The EVRC Vocoder is Varialbe Rate, the maximal encode rate is Skbps, its voice quality is closed to QCELP-13k, and has better ability of anti-disturbance. At the present time, EVRC is the best vocoder in the CDMA system when take into account both the voice quality and the encode rate.EVRC Vocoder is widely used in the Base Station of IS-95 and CDMA2000. Because our vocoder is mainly used in the Base Station, where the vocoder needs the excellent performance, so we select the TMS320C64xx , the most advanced DSPs of TI. The paper makes great efforts on the software optimization of EVRC vocoder. Based on the understanding of TMS320C64xx CPU structure, we do deeply-optimization on the loop which appear usually in voice signal processing, and this improve the utility ratio of CPU and the parallelity degree of CPU function cell. At the same time, we utilize the Bit-Exact test to test the fixed-point EVRC vocoder with the test vectors of TIA/EIA/IS-718, which improve the robustness of the vocoder.At last, our vocoder is compared with the vocoder Benchmark of Delphi company, with the result that the performance of our vocoder is better than Delphi’s obviously, thus make the vocoder has applied value.更多还原