为了提高非特定人汉语数码串连续语音识别系统的识别速度 ,对系统进行了细致的研究 ,有针对性地提出了一种新的快速识别算法 .通过对隐含马尔可夫模型输出概率密度函数运算的简化 ,以及采用结合段长信息的束搜索剪枝算法 ,在保证识别精度的情况下 ,使识别系统对不定长数码串平均识别时间从 0 .92 s减少为 0 .11s,而串识别正确率仅从 94.3%下降到 94.0 % ,大大地提高了系统的整体性能 In order to improve the recognition speed of non-specific Chinese digital string continuous speech recognition system, the system is studied in detail, and a new fast recognition algorithm is proposed in this paper.According to the output probability density function of implicit Markov model The simplification of operation and the algorithm of beam search and pruning which combined with the length information, the identification system can reduce the average identification time of indefinite long digital strings from 0.92 s to 0.11 s while identifying the accuracy, The correct rate dropped from 94.3% to 94.0%, greatly improving the overall system performance