摘要
针对语音识别中动态时间规整(DTW)对语音端点检测精确性过度依赖、识别时间长及识别效率低等问题。为提高语音识别精度和效率,采用改进型的蚁群算法来处理动态时间规划问题,核心是对基本蚁群算法采用自适应的挥发系数,动态信息素更新策略。用新的状态转移规则以及最优的蚂蚁参数选择等改进方法,使能在较短的时间内能寻找到最佳路径,提高执行效率。仿真实验分别测试了传统DTW算法和基于改进蚁群算法的DTW算法的识别率,结果表明,新算法的全局搜索能力、准确性都优于传统的DTW算法,能有效的提高语音识别系统的效率。
The result of the dynamic time warping（DTW） depends much on the accuracy of endpoint detection.The recognition time is too long and less efficient in the voice recognition.An ant colony optimization（ACO） algorithm is presented to solve the dynamic time warping problem.The core of this algorithm is using adaptive evaporation coefficient,dynamic pheromone update strategy,a new state transition rule and the ants with the optimal parameter,and so on.The algorithm can find a better route in a short time,and improve the performance.The simulation compared the traditional DTW with the DTW based on the improved ant colony algorithm,the results show that the new algorithm has better global search ability and accuratenss than the traditional ant colony algorithm and the traditional DTW.It can provide a better performance in the speech recognition rate.
出处
《计算机仿真》
CSCD
北大核心
2011年第5期402-405,409,共5页
Computer Simulation
关键词
蚁群算法
动态时间规划
优化
仿真
ACA
Dynamic time warping
Optimization
Simulation
