EEG source reconstruction is a challenging problem due to its ill-posed nature. In this research, we propose a multi-resolution version of the Multiple Sparse Prior (MSP) algorithm, such that the EEG inverse problem is solved in the low resolution space and the active regions are determined approximately then the source reconstruction is done in high resolution from the obtained source space. An advantage of this method is reducing the source space. Also, by locating the prior information in the active regions, the performance of the classic MSP algorithm improves and the higher model evidence is achieved because of importing the prior knowledge in to the problem. We use simulation to compare our proposed method with the classic MSP. We use the following performance measures to compare the methods: free energy, explained variance, relative root mean square error, and the spatial distance error. Our method outperforms the classic MSP in extracting the brain sources time series and their spatial maps.
展开▼
