本文介绍一种称之为谐差(Harmonic Deviations)声码器(HDV)的窄带语言编码器的发展、究究和试验结果。HDV编码器以LPC声码器为基础,除了LPC声码器所传输的数据外,它还传输谱差形式的附加信息,该谱差系指在选定频率组上语言谱和LPC全极点模型谱之间的差。在收端,则用这些谱差产生全极点综合滤波器的激励信号。本文描述并比较了以语言信号提取谱差并对谱差编码的若干种方法。为了将HDV编码器的比特率限制在2·4 kb/s,文中还讨论了包括正交变换和对数面积比的最小均方误差标量量化、两级矢量——标量量化以及可变帧率传输等。此外,文中还介绍了2·4kb/s HDV编码器的语言质量最佳化结果,研究了低噪声和噪声信道应用。指出所获得的HDV编码器在语言质量上比LPC声码器有了明显的提高。 This article describes the development, research, and experimental results of a narrowband speech coder called the Harmonic Deviations vocoder (HDV). Based on the LPC vocoder, the HDV encoder, in addition to the data transmitted by the LPC vocoder, transmits additional information in the form of spectral differences, which refer to the speech spectrum and the LPC all-pole model on the selected frequency group The difference between the spectra. At the receiving end, these spectra are used to generate the all-pole synthesis filter excitation signal. This article describes and compares several ways to extract spectral differences and encode spectral differences with speech signals. In order to limit the bit rate of HDV encoder to 2.4 kb / s, we discuss the minimum mean square error scalar quantization including orthogonal transform and logarithm area ratio, two-stage vector-scalar quantization and variable frame rate Transmission and so on. In addition, the language quality optimization of 2.4 kb / s HDV encoder is also introduced, and the application of low noise and noisy channel is studied. It points out that the obtained HDV encoder has a significant improvement in the quality of speech over the LPC vocoder.