Skip to Main Content
The accuracy of the source localisation using EEG measurements depends on the noise as well as the computational model, so the improvement of localisation performance by using a very realistic model is limited. In this reported work, the source localisation effects of various human head models with different geometrical deformation for low signal-to-noise ratio (SNR) measurement are investigated. Assuming the most dense mesh as a model without model mismatch, it was found that some error due to the model mismatch was tolerably small in the range of low SNRs less than 0 dB. This provides a cost-effective strategy of head model generation, depending on a given tolerable level of localisation error.