The Research Of Speech Codec Based On Reconfigurable Computing Technology

With the development of technology of internet and communication, VoIP technique is becoming to a hot researching domain. iLBC speech codec was especially designed for narrow band speech. The technique of frame-independent long-term predictive coding and packet loss concealment is adopted in the codec. The codec enables graceful speech quality degradation in the case of lost frames, which occurs in connection with lost or delayed IP packets. The advantages of the codec make it more and more widely used.By studying the iLBC speech codec ANSI C algorithm, a SoC solution based on reconfigurable computing technique was designed, which is easier for development, more flexible and easier to be updated compared with the DSP processor based solution.By analyzing the function of autocorrelation computation in the iLBC speech codec, a float point model of the function was established first and then a fixed point model was generated which was verified in the emulation model described by VHDL. However, the operation of fetch data from memory cost too much time, and because of which the whole performance of the model is severely declined.A parallel implementation of the fixed point model was created to avoid the disadvantage of the serial implementation, which decreases the need of fetching operation to five percent of serial implementation of fixed point model. A SoC solution with MicroBlaze as the embedded processor was established to verify both of the serial and parallel implementations of the fixed model.The system of iLBC speech codec is implemented on reconfigurable computing platform based on FPGA and the results of the experiment imply that the SoC solution proposed in the thesis is more efficient than DSP based solutions.