论文部分内容阅读
引言在16kb/s或更低速率的数字语言传输中,要想得到极高的语言质量是困难的.已经研制的2400-4800比特/秒的线性预测编码系统其可懂度是良好的.但由于提取激励参量(一般是指基音和有声音)的困难,质量上有些损失.而工作在8kb/s至16kb/s的系统通常是不能依赖于激励基音和有声音,但这不包括某些形式的差分脉码调制.这些系统虽因粒状量化噪声使话音质量有所损失,仍能很好地再现语言的激励函数.
Introduction It is difficult to achieve extremely high speech quality in digital speech transmission at a rate of 16 kb / s or lower. The 2400-4800 bit / s linear predictive coding system that has been developed has good intelligibility but due to Difficulties in extracting excitation parameters (generally referred to as pitch and voicing) are somewhat of a quality penalty, while systems operating between 8kb / s and 16kb / s generally can not rely on stimulus and speech, but this does not include some forms Of the differential pulse code modulation of these systems due to granular quantization noise so that the loss of voice quality, still well reproduce the language of the excitation function.