摘要
声音匹配识别技术是将被识别的目标声音对象和声音样本进行比对,得到目标声音和样本的一致性判断。本文通过对声音的特性分析,提出了一种高效的识别算法,采用声音特征参数提取、矢量化(Vector Quantization)技术、样本匹配等手段,克服了一般声音识别算法存在的占用CPU时间长、识别速度跟不上语速的弱点,提高了声音识别正确率。经过实际测试,识别准确率高达99%以上。
The definition of Sound Match Recognition Technology is comparing original sound sample with goal sound sample to identify their consistency. Analyzing the characteristics of sound, the paper comes out an effective recognition algorithm by using with feature pick -up technology, Vector quantization and discriminative matching methods and so on. The advantage of this algorithm is higher recognition accuracy by overcoming the obstacle of slow CPU respond speed. The recognition ratio reaches 99% during practical testing.
出处
《中国传媒大学学报（自然科学版）》
2007年第1期20-25,共6页
Journal of Communication University of China：Science and Technology
关键词
声音识别
矢量化
匹配算法
DTW
sound recognition
VQ
discriminative matching
DTW
