论文部分内容阅读
本文描述了一个用于传输语言以6400比特/秒速率的四阶自适应预测编码器(APC)之实时计算机的执行过程。为了降低中央处理器的要求,终端采用平均量差函数(AMDF)来提取基音,假定输入数据稳定时,计算预测器系数和反射系数。使用16比特定点运算处理,编码器在判定音韵试验时获得的可懂度可达百分之八十七,甚至存在声学背景噪声时产生的语言、其话音自然度也是可接受的。
This article describes the implementation of a real-time computer that uses a fourth-order adaptive predictive coder (APC) for transmitting speech at 6400 bps. In order to reduce CPU requirements, the terminal extracts the pitch using the Mean Difference Function (AMDF), which calculates the predictor coefficients and the reflection coefficients assuming the input data is stable. With 16-bit fixed-point arithmetic, the encoder achieves an intelligibility of 87% when deciding on phonological tests, even in the presence of acoustic background noise, whose speech naturalness is also acceptable.