It has been verified that the ICA can isolate sources from multichannel magnetoencephalography (MEG) signals. Based on the route of constrained ICA (cICA), this paper achieves a new solution of MEG inverse problem called functional source separation (FSS) by adding a functional constraint to the cost function of a basic ICA model. Source activity is obtained by applying this method to one MEG signal dataset under a selfpaced finger tapping task. The result is proved effective by calculating correlation coefficients between the weight vectors of function source separation method and the spatial filter coefficients of SAM method. It is found that finger tapping related functional source was localized in motor cortex of precentral gyrus. At the same time, the temporal and frequency information provided by FSS method could be a basis of exploring cortical control timing mechanisms associated with finger movements and extracting time frequency characteristics of the functional source.