Research Of Low Bit Rate Speech Coding Based On DSP

In recent years,with the information society and communication technology rapid development, With the increase requirement of communication,the communication resources become more and more rarity.Now speech communication is still an important operation,thus,speech compression appeals more people to regard on it for a long time.High quality synthetical speech on low bit rate is aspiring.these conditions of low rate speech coding put forward higher requirements.Among the mang low-rate coding method, Mixed Exciation Linear Prediction codec with low bit rate,low delay,low complexity,etc. So MELP algorithm is one of the very best low rate speech coding.It is encoded in the original of LPC,the introduction of multi-band mixed excitation,aperiodic pulses,residual harmonic processing,adaptive spectral enhancement techniques,and pulse shaping filtering.Adopted a new mechanism is more consistent with human voice pronunciation model to generate synthetic speech,synthetic speech to match the original voice of the high degree of order to better achieve the low bit rate speech coding.In this paper, the thesis mainly more in depth study the theory of MELP voice coding algorithms, and the extraction of the pitch improvement made,Besides that,simulation analysis of algorithm is presented by using Matlab, This thesis alse wrote the C program for the MELP algorithm,So based on the C program,this thesis implements the MELP algorithm on the TMS320VC5509 DSP of Texas Instruments TI company,Make use of TLV320AT23 pick the voice signal ,and by means of McBSP will be pick voice signal that is transmitted to the TMS320VC5509,at last,implemented on the TMS320VC5509 voice MELP algorithm. and the results of the simulation.Real-time implementation of algorithm for computer large problems arise,the code proposed algorithm optimization method.