摘要 |
<p>The invention discloses a coding/decoding method based on templet matching and multi-distinguishability analysis. An inputted audio signal is divided into a slow-change signal and a quick-change signal according to the signal performance. If said audio signal is the quick-change signal, time-frequency distinguishability of the multi-distinguishability analysis will be determined, and then a time point and a frequency range which need to make templete matching according to the time-frequency distinguishability are selected. Using coefficients on the selected time-point along the direction of frequency, linear prediction is made and whether the multi-distinguishability analysis templet matching is needed is judged. If necessary multi-distinguishability analysis templet matching step and residual error coding step are performed. Border information is written into stream of bits, and then the compressed stream of bits is output. The invention uses the templet matching method. For these coefficients which are in collection of energy direction, parameters which represent respective templet are obtained and quantization encoding is performed in accordance with the parameters. For residual error, further perception quantization encoding is performed. For the quick-change signal, higher coding gain can be obtained, and pre echo phenomenon can be effectively prevented.</p> |