摘要
文章首先简单介绍了压缩感知(CS)理论框架,然后根据语音信号小波变换系数的特点,提出了改进的压缩感知算法,对高频系数进行压缩处理,低频系数不变。采用基追踪算法重构出高频系数,再利用小波反变换得到原始语音信号。实验结果表明,在相同的测量点数下,本文的算法比原有CS算法在重构语音的信噪比和MOS分上都有较大的提升。
The CS framework is introduced briefly,and then according the features of the wavelet transform coefficients of speech signals,an improved CS algorithm is proposed,which preserves the low frequency wavelet transform coefficients but compresses the high frequency wavelet transform coefficients of the speech.Finally,the high frequency wavelet transform coefficients are recovered by using basis pursuit（BP） algorithm,and then the reconstruction of the speech signal can be achieved by the inverse wavelet transform.Compared with the original CS algorithm,the proposed algorithm has significant improvement on both SNR and the MOS scores of the reconstructed speech at the same measurement number.
出处
《电子技术（上海）》
2011年第7期10-11,共2页
Electronic Technology
基金
国家自然科学基金(60971129)资助
关键词
压缩感知
小波变换
稀疏性
基追踪
compressed sensing
wavelet transform
sparsity
basis pursuit
