摘要
在声音识别系统中,特征参数的获取对声音识别和训练有着重要的影响;MFCC算法作为典型的声音特征参数提取方法,性能稳定,识别率高;针对MFCC算法存在较大计算量的情况,提出一种改进的特征参数提取算法MFCC_E;相比于标准的MFCC算法,MFCC_E算法减少了约50%的运算量,并且易于硬件实现;实验结果表明,MFCC_E算法与MFCC算法的识别率大致相同,而计算复杂度却小很多。
In audio recognition system, the acquisition of characteristic parameters has important influence on audio recognition and training. MFCC algorithm, as a typical audio characteristic parameter extraction method, has stable performance and high recognition rate. According to the situation that MFCC algorithm has larger amount of computation, a kind of improved characteristic parameter extraction algorithm, MFCC E,was pointed out. Compared with standard MFCC algorithm, MFCC_E algorithm reduced about 50% computation amount and was easily realized on hardware. Experiment results show that the recognition rate of MFCC_E algorithm is approximately the same as MFCC algorithm but the computation difficulty of MFCC_E is largely smaller than that of MFCC algorithm.
出处
《重庆工商大学学报（自然科学版）》
2012年第2期52-57,共6页
Journal of Chongqing Technology and Business University:Natural Science Edition
基金
重庆市科委重大攻关项目(CSTC
2010AA2036)
重庆市教委项目(KJ100709)
