Proposed in this paper is a new fixed-point blind extraction algorithm of μ-rhythm electroencephalogramsignals based on high-order statistics,which takes into consideration the super-Guassian characteristic and non-symmetricdistribution of the signals. Then,the local stability and convergence of the proposed algorithm are discussed,and the corresponding conditions are determined. Finally,the algorithm is compared with the FastICA algorithmthrough a case study on simulated and clinical μ-rhythm electroencephalogram signals,with a higher effectivenessof it being revealed.