低速语音编码算法的设计和实现-通信与信息系统专业论文.docxVIP

摘要 摘要 摘要 在现代社会中，随着数字通信的快速发展和广泛应用，低速语音编码算法在 窄带通信，特别是军事或安全通信等领域中发挥着重要的作用，并且也广泛应用 于民用通信中。混合激励线性预测 MELP 是一种性能比较好的低速语音编码算法。 MELP 在线性预测模型的基础上作出改进，使用了非周期脉冲、混合激励、自适 应谱增强、脉冲整形滤波和傅里叶级数幅值；并且结合了多带的思想，具有多带 激励和线性预测编码的优点。 本文以 MELP 编解码算法为模型，设计和实现了 1.2kbps 的低速语音编解码 算法。由于可用来表征语音参数的比特数比较少，1.2kbps 的语音编码算法很难达 到很好的质量，所以本文使用两个子帧构成一个超级帧。每超帧内分别对两个子 帧的参数进行提取，其中带通清浊判决、增益和线谱频率仍采用单帧量化，而基 音周期和傅里叶级数幅值两帧联合进行矢量量化编码。最后将 1.2kbps 低速语音 编码算法移植到 TI C6000 系列的 DSP 平台上，并对其进行优化。 测试结果表明，设计的 1.2kbps 语音编解码算法的合成语音的 PESQ 值比 2.4kbps MELP 算法合成语音的 PESQ 值降低约 0.17。说明本文设计算法合成语音 的质量接近 MELP 的合成语音。在 DSP 平台上对算法进行实现，优化后算法复杂 度降低到优化前的 10%以下，为工程应用提供了良好的基础。 关键词：信源编码 低速语音编码 MELP 超级帧 矢量量化 Abstr Abstract Abstract With the rapid development and the widespread applications of the digital communication, the low bit rate speech coding plays an important role in narrow band communications, especially in military or secure communications, and in civilian use as well. Mixed excitation linear prediction (MELP) vocoder is one of the most potential algorithm in modern low bit rate speech coding. MELP is based on the model of LPC vocoder and includes five additional features which are mixed excitation, aperiodic pulses, pulse dispersion, adaptive spectral enhancement and Fourier magnitudes scaling of the voiced excitation. Based on the model of MELP, a very low bit rate speech coding algorithm working at 1.2kbps is designed and implemented in the thesis. Firstly, because of the low bit rate speech coders at 1.2kbps can hardly get good performance for the lack of bits to accurate repersent the parameters of speech signals, a superframe that composes two subframes is used. Secondly, for every subframe, the parameters of speech signals are analyzed and extracted. Thirdly, subband voiced/unvoiced judgment, gain and linear spectral frequency (LSF) are quantized in one subframe, and joint vecter quantization algorithms are applied to pitch and amplitude value of Fourier series of the residual signals in two subframes. F