摘要: 独立分量分析法在分离含有背景噪声的混合语音时效果不理想。为此,将独立分量分析算法与卡尔曼滤波相结合,对语音进行降噪处理,采用FastICA算法对含噪语音进行分离,分离速率高于Informax算法,能够获得较清晰的语音文件。通过仿真验证了该方法的可行性和有效性。
关键词:
独立分量分析,
背景噪声,
卡尔曼滤波,
语音分离,
FastICA算法
Abstract: Because of background noisy, there is a problem of directly separating the mixed-signal observed usually can not achieve good results. The basic principle and algorithms of independent component analysis is illustrated. Considering characteristic of Kalman filtering, the speeches under background noisy are denoisied; and speeches are separated by the FastICA algorithms. Simulation results indicate the availability and efficiency of this method.
Key words:
Independent Component Analysis(ICA),
background noise,
Kalman filtering,
speech separation,
FastICA algorithm
中图分类号:
云晓花, 景新幸. 背景噪声下的语音信号分离[J]. 计算机工程, 2011, 37(23): 181-182,185.
YUN Xiao-Hua, JING Xin-Nie. Speech Signal Separation Under Background Noise[J]. Computer Engineering, 2011, 37(23): 181-182,185.