The Implementation And Improvement Of MELP Speech Coding Algorithm

Along with the rapid development and the widespread applications of the digital communication in the social life, the low rate speech coding played an important role in the fields such as network communication, military communication, secure communication and satellite communication, became the key technology of speech coding. Among the speech coding which were standardized, MELP(Mixed Excitation Linear Prediction) based on the model of LPC Vocoder, made the 5 parts of improvements, and combined with the algorithm of MBE, made it worked on the 2.4kb/s, simulated the characteristics of natural speech in order to synthesize the sound with high-quality. So MELP became the hot technology in the low rate speech coding, been standardized by U.S. federal standards.Researching speech coding Algorithm, for applied to the project to play its role in encoding and decoding speech signals. In the process of engineering, because the computational complexity of speech coding is often very large, always had limitation in running time, and limited by different application platform, so chose the appropriate application platform ,made the program running on the platform efficiently became an indispensable link in the engineering applicationThe thesis researched on technology of low bit rate speech coding mainly in MELP, solved the problem of optimized the program of MELP on DSP platform. The thesis also improved the algorithm of MELP for lower the rate of coding. The main contents include:1) Base on the theory of speech coding, the thesis researched the theoretical basis of low bit rate speech coding. And studied the currently advanced low bit rate speech coding algorithms from the algorithm system of low bit rate speech coding.2) The thesis researched the MELP speech coding algorithm detailedly, and analyzed the role of various parameters, calculation methods, the principle of quantification, and methods of speech synthesis. Used the C language programmed the MELP speech coding algorithm on Microsoft Visual Studio 2003, assessed the quality of synthesized speech through P.862(PESQ) which be recommended by ITU, And analyzed the waveform and the spectrogram of the synthesized speech.3) In the context of engineering application, the thesis implemented the MELP speech coding algorithm on DSP platform, and optimized the program on DSP platform for reducing the operating time of program. During the optimization process, the thesis achieved the effect of optimization simplify and rapidly, by using different methods which includes compiler tools, small function inline expansion and intrinsics and so on. The results would be ideal to meet the needs of engineering applications.4) In order to appliance the MELP speech coding algorithm more widely, the thesis researched the improvement of MELP speech coding algorithm, proposed improved through increasing the length of each frame, reducing the numbers of quantification bit of bandpass voicing by the methods of classification, researched the quantification of LSF, reduce the number of quantification bit by using the inter-frame correlation, and made two kinds of improved algorithm of MELP. By using the C language program the improved algorithms on Microsoft Visual Studio 2003, analyzed the results. The results of synthesized speech showed two kinds of improved algorithm produced the high quality synthesized speech by working on the coding rate of 1.2kb/s.