Extraction of EMG signal buried in RMI noise

The electromyographic signal (EMG) describes the muscle activity. But, when acquired under special conditions, such as functional Resonance Magnetic Images (fMRI), an important amount of noise, hide completely the useful signal. The purpose of this paper is to analyze fMRI noise characteristics and to develop an approach to reduce it. The proposed idea is based on spectral subtraction of the noise spectrum estimate over the entire EMG spectrum. Evaluation of the technique performances reveal that spectral subtraction is a viable artifact reduction tool that should be used with EMG collected during fMRI data acquisition.