突发业务下共享存储区型ATM交换单元的仿真

Apply the Knowledge from Psychophysics of Hearing to Isolated Digits Speech Recognition

通过建立共享存储区型ＡＴＭ交换单元的仿真模型，研究了突发业务下吞吐量、平均突发长度、共享区长度与信元丢失及延迟的关系，发现不同输入循环方式将引起信元丢失在输入端的分布差异．还发现，吞吐量中等或端口流量特性不同时，部分共享方式可带来好的效果．

Perceptual linear predictive cepstral parameter (PLPC) is proposed based on the knowledge from psychophysics of hearing and is used in isolated digits speech recognition.PLPC is compared with linear predictive cepstral parameter (LPCC), the speech recognition rate of using eight order PLPC is higher than that of twelve order LPCC, Multi featuer parameter is also used based on PLPC, the recognition rate of speaker independent is approximately 97 %