中文说明：应用背景语音编码是现代语音的重要支撑技术之一通信系统，和非常低的语音编码算法的研究是一个重要的语音编码区的研究方向。本文涉及的关键技术的MELPe语音编码算法。在本文中，我试着完成下面的工作：（1）学习和实现语音增强算法。这种方法转换音频频域噪声的语音信号，然后根据树皮临界带宽，它将信号分为一组不互相搭接的频率段。在每一个频率段，根据信噪比得到增量，并根据不同的噪声和语音的不同特点是不同的频率功率，它使噪声的语音的幅度谱乘以增量为了抑制噪声，提高语音的主观感觉质量。（2）研究语音参数的估计算法。这些参数包括基音周期、增益、语音/清音化决策、傅里叶振幅、线谱对，等。1200bps和600bps算法，对平滑算法的研究我的论文间距周期。（3）我研究的重要内容是参数量化算法。在600bps算法，为了进一步降低超帧位，采用分类和图形化的参数定量方案。该方案不仅是多种超级帧，而且还分类模式根据相邻帧的差异。在同一个模式，但不同的模式或相同分类不同模式，我们采用不同的量化方案。在对多帧基音周期参数量化，采用定量方法根据基本频率跟踪，因此，使用更少的比特，以进一步量化的基本频率。（4）提出了1200bps、2400bps速率转换算法。通过设置缓冲区长度，我们可以实现2个速度的约定。（5）本文研究了一种新的混合励磁发电方法。这种方法提高了MELP算法的激励生成方法和采用混合边界的变化根据带通声强度的方法。（6）我们实现2400bps / 1200bps / 600bps的编码和解码算法。实验结果结果表明三种算法都具有较高的清晰度和可懂度。2400bps算法具有较高的自然。关键技术低比特率语音编码；语音参数估计；语音参数量化；语音增强；比特率转换
English Description：
Application background speech coding is one of the important supporting technologies of modern speech. The research of communication system and very low speech coding algorithm is an important research direction of speech coding area. This paper deals with the key technology of melpe speech coding algorithm. In this paper, I try to complete the following work: (1) learning and implementing speech enhancement algorithm. This method converts the audio frequency domain noise speech signal, and then divides the signal into a group of non overlapping frequency bands according to the critical bandwidth. In each frequency band, the increment is obtained according to the signal-to-noise ratio, and the frequency power is different according to the different characteristics of different noise and speech. It multiplies the amplitude spectrum of noise speech by the increment, in order to suppress noise and improve the subjective quality of speech. (2) The estimation algorithm of speech parameters is studied. These parameters include pitch period, gain, speech / unvoiced decision, Fourier amplitude, line spectrum pair, etc. 1200BPS and 600bps algorithm, the research of smoothing algorithm, my paper interval cycle. (3) The important content of my research is the parameter quantization algorithm. In the 600bps algorithm, in order to further reduce the superframe bit, the classification and graphical parameter quantitative scheme is adopted. This scheme is not only a variety of super frames, but also classification mode according to the difference of adjacent frames. In the same pattern, but different patterns or different patterns of the same classification, we use different quantitative schemes. In the quantization of multi frame pitch period parameters, the quantitative method is used to track according to the basic frequency. Therefore, fewer bits are used to further quantify the basic frequency. (4) The rate conversion algorithms of 1200BPS and 2400bps are proposed. By setting the buffer length, we can achieve two speed conventions. (5) This paper studies a new hybrid excitation generation method. This method improves the excitation generation method of MELP algorithm and the method using mixed boundary changes according to the band-pass sound intensity. (6) We implement 2400bps / 1200BPS / 600bps encoding and decoding algorithms. Experimental results show that the three algorithms have high clarity and intelligibility. 2400bps algorithm has high natural frequency. Key technologies: low bit rate speech coding; speech parameter estimation; speech parameter quantization; speech enhancement; bit rate conversion