摘要
为了解决传统DTW算法准确度和效率不高的问题,提出了一种基于图像配准方法的语音特征匹配算法.该方法将MFCC参数映射为二值图像,并通过引入图像配准的方法进行模板匹配,达到了语音特征匹配的目的.实验结果表明,与传统的DTW算法相比,该方法的准确率、召回率和算法执行效率有了明显的提高.
To overcome the low accuracy and efficiency of the traditional DTW algorithm used in speech feature matching, an approach employing the image registration algorithm is proposed. The speech feature matching process was implemented by mapping MFCC coefficients to the binary image, and introducing the image registration algorithm to the template matching. The experimental result shows that, comparing with the traditional DTW algorithm, the proposed algorithm achieves a better performance in precision rate, recall rate and the cost of computation.
出处
《哈尔滨工业大学学报》
EI
CAS
CSCD
北大核心
2008年第7期1152-1155,共4页
Journal of Harbin Institute of Technology
基金
吉林省科技发展计划国际合作项目(20050703-1)
关键词
DTW
模板匹配
图像配准
语音识别
DTW
template matching
image registration
speech recognition
