Research Of Low Bit Rate Speech Compressing Based On DSP

With the rapid development of information and communication technologies, voice coder has turns to be an important part of digital voice communication technology. Under the rapid development of military communications, mobile communications and internet, bandwidth resource is becoming very expensive and in short supply. Therefore, low bit rate speech coding technology has been an important research direction of speech communication.2：4kbps mixed excitation linear prediction (MELP) is a modern low bit rate speech coding algorithm with high performance. MELP is based on the traditional model of LPC vocoder and with five new technologies, which are multi-band mixed excitation, aperiodic pulses, adaptive spectral enhancement, pulse dispersion filtering, and residual harmonic treatment. Therefore it can make the synthesized voice more naturally and the bit rate lower.Firstly, the principles and the process of MELP are analyzed deeply, then Fourier magnitude of residual signal is extracted. From the concrete analysis, we can find that in the process of extracting Fourier magnitude the phase information of residual is ignored. Therefore the quadratic linear prediction coding and decoding algorithm is proposed in this paper, and the performance of the algorithm is verified by simulation experiments. Then the performance of the algorithm as the length of processing frame changes is studied. Finally the algorithm is implemented in C language, then transplanted to DSP with using software simulation function of CCS integrated development environment.The simulation results have shown that the quality of quadratic linear prediction coding algorithms is slight better than that of MELP in the same bit rate. When the length of processing frame reduced to20ms, namely160sampling points per frame, the performance of quadratic linear prediction coding algorithms improved. Finally the algorithm is implemented in C language and applied to the TMS320C55X series DSP of CCS simulation environment, which lays the foundation for the further application in the engineering practice.