Abstract: The invention discloses a sparse Bayesian learning (SBL)-based SSR brain source localization method. An SSR record is divided into multiple data segments, frequency-domain information of the data segments is extracted through FFT, and a data matrix is constructed. An automatic iteration stop condition and initial values of a sparse support vector and a spontaneous electroencephalography (EEG)-electrical noise joint power vector are set. The posteriori mean and covariance of SSR components are iteratively estimated and the sparse support vector and the spontaneous EEG-electrical noise joint power vector are updated accordingly. When the iteration ends, the ultimate sparse support vector is used to give a source localization result. An SSR source localization problem is modeled in the frequency domain, the joint sparsity of signals in multiple data segments is involved, and a brain source localization method applicable to various SSRs is given in an SBL framework.

Abstract: The invention discloses a sparse Bayesian learning (SBL)-based SSR brain source localization method. An SSR record is divided into multiple data segments, frequency-domain information of the data segments is extracted through FFT, and a data matrix is constructed. An automatic iteration stop condition and initial values of a sparse support vector and a spontaneous electroencephalography (EEG)-electrical noise joint power vector are set. The posteriori mean and covariance of SSR components are iteratively estimated and the sparse support vector and the spontaneous EEG-electrical noise joint power vector are updated accordingly. When the iteration ends, the ultimate sparse support vector is used to give a source localization result. An SSR source localization problem is modeled in the frequency domain, the joint sparsity of signals in multiple data segments is involved, and a brain source localization method applicable to various SSRs is given in an SBL framework.