The functional connectivity between different brain regions is of importance for human brain to accomplish cognitive tasks

and it may appear as correlated temporal behavior of neural activity.It will help to investigate the brain underlying mechanism to noninvasively localize coherent sources which underlie the scalp EEG recordings.Classical MUSIC(Multiple Signal Classification) method has difficulty on localizing coherent sources.This paper presented a class of MUSIC for coherent sources localization in transformation data space.At first

the coherent source region is coarsely estimated by a prior knowledge or other mapping methods.And then

a transformation matrix designed to suppress the source activity is constructed.At last

using the transformation matrix to project the scalp EEG recordings into a new transformation data space

where the coherent sources can be localized by classical MUSIC method.Computer simulations reveal that under different levels of noise

in contrast to other coherent source localization algorithms

the proposed method has rather less mean localization bias and run much faster.