The present invention provides a method for identifying images of brain function. In the beginning, choosing one of the brain data collected by multichannel scalp EEG/MEG, and using a mode decomposition method to obtain a plurality of intrinsic mode functions for each brain data, transforming the intrinsic mode functions (IMFs) in the same frequency scale into a plurality of source IMFs across the cerebral cortex by a source reconstruction algorithm, and classifying each source IMF in the same frequency scale into a plurality of frequency regions corresponding to the different brain sites. Then, repeatedly choosing a source IMF, and obtaining an amplitude envelope line through each absolution value of the source IMF. Further to obtain a plurality of source first-layer amplitude IMFs decomposed from the function of the amplitude envelope line by the mode decomposition method. Until obtaining the source first-layer amplitude IMFs from each source IMF, classifying each source first-layer amplitude IMF in the same amplitude frequency scale into a plurality of amplitude frequency regions corresponding to the different brain sites. In the end, a brain amplitude modulation spectrum is provided for analyzing the relationship between each frequency region and each amplitude frequency region.
