摘要
提出了一种基于小波变换和HMM模型的ARMA新模型参数 ,并将它用于A…N的英文字母的识别。小波变换可以在高频提供高的频率分辨 ,在低频提供高的时间分辨率 ,而ARMA模型则可以改善LPC模型中没有零点的不足。实验结果表明 ,2个零点 ,10个极点的ARMA对字母C的识别准确性明显提高。
Some parameters based on wavelet transform and ARMA Model are presented. Wavelet transform provides a high frequency resolution, and the ARMA model is more powerful due to its including of the zeros pole in the model. The experimental result of alphabet of A to N from National Institute of Standard Technology (NIST) database is given. The error rate has been improved, especially C.
出处
《重庆大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2001年第4期61-63,共3页
Journal of Chongqing University
