Study And Implementation Of MELP Speech Coding Algorithm Based On DSP

With the continuous development of information society, media and memory resources become more and more valuable, so compressing the transmitting bandwith of speech signal has been a target to be pursued.A new mixed excitation linear prediction (MELP) codec working at very low bit rate was presented. This codec utilizes the advantages of both linear prediction coding(LPC) algorithm and multi-band excitation (MBE) coding algorithm. In order to get natural speech with good quality, some new features of the algorithm and quantization scheme were employed including mixed excitation, aperiodic pulses, adaptive spectral enhancement, pulse dispersion, and Fourier magnitude modeling. The synthetical speech of MELP is legible and natural, which has a high signal to noise ratio and compression ratio.In this dissertation, the fundamentals of MELP was studied and some of the formulas were validated. The coding and decoding simulation is done on PC by C language. Before being transplanted to TMS320VC5510 DSP, the programme needs to be optimized under CCS tool in order to improve the running efficiency of the code. At last, the programmes of MELP optimized were realization on TMS320VC5510 DSK. It is proved that the MELP algorithm is an excellent low bit speech coding method and has high practicality.