摘要
首先对语音和背景音乐以及两路语音信号两种情况下的混合语音信号进行盲源分离,得到目标信号;其次对目标信号进行EMD分解得到一组固有模态函数,再求出各固有模态函数的能量值Ei;再以若干组能量值为训练样本,通过LS-SVM对相应的语音信号进行分类研究。通过实验验证,该方法在语音信号研究方面能取得较好的分离效果。
Firstly, blind source separation algorithm is applicated to two cases, one of which is mixed with speech and background music, the other is mixed with two voice signal. Then, the EMD algorithm is applicated to the target signal, which can obtain a set of intrinsic mode functions and Ei of each intrinsic mode function of energy. Finally, with several groups of energy as the training sample, the corresponding voice signal is studied through the LS -SVM classification. Experiments verify that the method is efficacious to audio separation.
出处
《电声技术》
2014年第6期48-52,共5页
Audio Engineering
基金
河南城建学院2013年度科学研究基金项目(2013JYB010)
