Recently , the study of the low bit-rates(4800-9600bps) speech coders mainly cenlres on aclass of analysis-by-synthesis linear predictive speech codersi which includes many well-known tech-niques. In this class the co...Recently , the study of the low bit-rates(4800-9600bps) speech coders mainly cenlres on aclass of analysis-by-synthesis linear predictive speech codersi which includes many well-known tech-niques. In this class the coders all have near toll qualzty , but the high complexity gets rid of their mod-est hardware implementation. In this paper, an enhanced self-excited vocoder with an harmonic noiseweighting filter(HSEV) is presented in which several techniques for zmproving the speech qualiiy beused. An harmonic noise wdghted adaptive code book which is similar to a closed loop lonAr-term pre-diction filter is used to excite the system. For decreasing the computation, an hybrid open/clesed-loopsearch procedure is used. The noise wezghting filter is a cascade of a traditional short-term spectralweighting filter and an harmonic noise weighting filter. Finally the decoded speech signal's quality ismere improved by an adaptive post filter. The overall algorithm is very simple, but us decoded speechis sound. This speech coder can be implemented in a single DSP such as TMS320C25 with an integralecho canceller展开更多
文摘Recently , the study of the low bit-rates(4800-9600bps) speech coders mainly cenlres on aclass of analysis-by-synthesis linear predictive speech codersi which includes many well-known tech-niques. In this class the coders all have near toll qualzty , but the high complexity gets rid of their mod-est hardware implementation. In this paper, an enhanced self-excited vocoder with an harmonic noiseweighting filter(HSEV) is presented in which several techniques for zmproving the speech qualiiy beused. An harmonic noise wdghted adaptive code book which is similar to a closed loop lonAr-term pre-diction filter is used to excite the system. For decreasing the computation, an hybrid open/clesed-loopsearch procedure is used. The noise wezghting filter is a cascade of a traditional short-term spectralweighting filter and an harmonic noise weighting filter. Finally the decoded speech signal's quality ismere improved by an adaptive post filter. The overall algorithm is very simple, but us decoded speechis sound. This speech coder can be implemented in a single DSP such as TMS320C25 with an integralecho canceller
