一种分离噪声盲源信号的FastICA改进算法-《集美大学学报（自然科学版）》

文章信息/Info

Title:: A New Improved FastICA Algorithm in Separating Noise Blind Source Signal

作者:: 方艺鹏1; 2; 周海峰1; 2; 张恩来3; 林开荣4; 郑东强4; 林忠华4; (1.集美大学轮机工程学院,福建厦门 361021；2.福建省船舶与海洋工程重点实验室,福建厦门 361021；3.集美大学诚毅学院,福建厦门 361021；4.集美大学海洋装备与机械工程学院，福建厦门 361021)

Author(s):: FANG Yipeng1; 2; ZHOU Haifeng1; 2; ZHANG Enlai3; LIN Kairong4; ZHENG Dongqiang4; LIN Zhonghua4; (1.School of Marine Engineering,Jimei University,Xiamen 361021,China;2.Fujian Province Key Laboratory of Naval Architecture and Ocean Engineering,Xiamen 361021,China;3.Chengyi University College,Jimei University,Xiamen 361021,China;4.School of Marine Equipment and Mechanical Engineering,Jimei University,Xiamen 361021,China)

Keywords:: noise source identification; FastICA; maximum similarity criterion; optimization algorithm

摘要:: 大功率机械所产生的非平稳噪声源，用FastICA算法可以有效地分离出噪声信号，但噪声源识别的准确性较低，为此提出了一种分离噪声源的FastICA改进算法。新方法将最大相似准则、优化算法和快速傅里叶变换相结合，对分离信号的不确定性进行了有效校正。经仿真实验验证，新方法能够完整地还原仿真波形信号，表明该改进算法确实可行且准确性较高。

Abstract:: Using FastICA algorithm may effectively separate noise signals from non-stationary noise sources produced by high-power machinery,although the accuracy is relatively low in identifying the noise sources.Therefore,a noise source separation method based on the improved FastICA algorithm is proposed.By combining the maximum similarity criterion,optimization algorithm and fast Fourier transformation,the uncertainty of the signal separation is effectively corrected. Simulation experiments prove that the method can completely restore the simulation waveform signal,which shows that the improved algorithm is feasible and accurate.