The real-time realization of filtering of speech with DSP TMS320VC5416 Chip

For the recognition of speech, the effect of speech preprocessing directly influences speech recognition. The filtering of speech is an important part of speech preprocessing, so the study of real-time realization of filtering of speech is very essential. This paper presents design of minimum speech system based on fixed-point DSP TMS320VC5416 chip. We use this system to realize Liner phase FIR filtering of speech in real time, liner phase FIR filter of Hamming function is used to filtering of speech for reducing Gibbs Effect, fixed-point arithmetic is adopted in filtering algorithm for reducing time of filtering of speech and ameliorates the real time. The system is made through connecting interface module of speech codec ASIC TLV320AIC23 to DSP teaching device. Since TLV320AIC23 Chip can directly connect to DSP Chip, hardware design is simple. When working frequency of TMS320VC5416 is set 160MHz, sampling frequency is 44.1kHz, pass band range of filter is 600Hz~1000Hz, the length of filter equal to 33, the simulation results show that running time of filtering is less than sampling time of one speech data, the filter can meet system´s real-time. The simulation waves show that filtering feature can meet the design requirements. Since the system makes full use of the DSP teaching device, the hardware design of system is simple, In addition, it combines with speech processing, it is as a good case of DSP or speech processing curriculum.