Realization of a multirate speech codec utilizing spectrum peak emphasis
Tetsu Taguchi · The Journal of the Acoustical Society of America · 1988
One of the most promising means of realizing multirate speech coding is multipulse excited linear predictive coding (MPELPC) adopting an auditory weighting filter (AWF). This AWF effectively shapes quantization noise to be masked by the speech signal and ameliorates speech quality at approximately 8 kb/s or more. However, below this rate, the AWF is not effective because the SNR is so low that the signals scarcely mask the noise. When the SNR is rather low, only the high-power frequency components of speech signals (whose levels are higher than those of the noise) should be encoded. In order to emphasize these components effectively upon encoding, a spectrum peak emphasis (SPE), which is performed by a prefilter and utilized in MPELPC, has been proposed. Below 8 kb/s, it has been shown by computer simulation that compared to the AWF, the SPE improves SNR at least by 2.2 dB. Subsequently, a multirate code from 16 to 4.8 kb/s, basically utilizing the AWF but with the SPE below 8 kb/s, has been realized on a DSP, NECμPD77230. This coder is applicable to mobile satellite communication.