Comparison of covariance-based and waveform-based subtraction methods in removing the interference from button-pressing finger movements.

A covariance matrix-based subtraction method has recently been proposed to remove interference using two MEG measurements: The first has both target and interfering activities and the second only has the interference. This paper compared covariance matrix-based subtraction with conventional waveform-based subtraction, which requires that the waveforms of interference be equal at every time point between the two measurements. Our analysis showed that covariance-subtraction only requires that the time-average of the squared intensity of interference be equal between the two measurements. As a result, the method is still effective when the onset of interference differs or even their measured waveforms differ between the two measurements. The covariance- and waveform-subtraction methods were both applied to remove the interference caused by response-button-pressing finger movements in auditory-evoked MEG measurements. The results of this application demonstrated the superiority of the covariance-subtraction method over the conventional waveform-subtraction method.