论文部分内容阅读
本文介绍了几种用于ADPCM编码系统的自适应技术及其原理、算法和设计,并通过微机模拟作了比较。就某些特定的语音帧而言,前馈自适应量化器与固定量化器相比较,大约可得到 3～6dB的分段(帧)信噪比增益。对于自适应预测器,若采用基于MMSE准则的自相关法确定最佳预测系数(包括反向自适应),系数是稳定的。但如果用梯度法确定最佳预测系数,则需要仔细地选择公式中的一些参数,否则,对某些特定的语音帧,可能使系统不稳定。
This paper introduces several adaptive techniques used in ADPCM coding system and its principles, algorithms and design, and compared by computer simulation. For some specific speech frames, the feedforward adaptive quantizer can obtain about 3 ~ 6dB of sub-frame (signal to noise ratio) gains compared with the fixed quantizer. For adaptive predictor, the coefficient is stable if the best prediction coefficient (including inverse adaptive) is determined by the autocorrelation method based on MMSE criterion. However, if you use the gradient method to determine the best prediction coefficient, you need to carefully select some of the parameters in the formula, otherwise the system may be unstable for some specific speech frames.