论文部分内容阅读
本文用较为简单的最速下降法设计了一种固定系数预测矢量量化器,其性能与应用普通的LPC法设计的预测矢量量化器基本相同。计算机模拟结果表明:在矢量维数K=8,数码率为8kbit/s的情况下,用最速下降法设计的预测矢量量化器与基本矢量量化器相比,信噪比和量化语音质量均有明显改善。文中介绍了用最速下降法设计预测矢量量化器的具体方法,并就预测矢量量化器与基本矢量量化器两者在复杂度、信噪比和量化波形等方面进行了比较。
In this paper, a simpler steepest descent method is used to design a fixed-coefficient predictive vector quantizer. Its performance is basically the same as that of the predictive vector quantizer designed by the ordinary LPC method. Computer simulation results show that when the vector dimension is K = 8 and the digital rate is 8 kbit / s, the prediction vector quantizer designed by the steepest descent method has better signal-to-noise ratio and quantized speech quality compared with the basic vector quantizer Significant improvement. In this paper, the method of steepest descent method for predicting vector quantizer is introduced, and the comparison between the prediction vector quantizer and the basic vector quantizer is carried out in terms of complexity, SNR and quantization waveform.