Muscles movement intention detection from EEG using movement related cortical potentials (MRCPs)

In brain-computer interface (BCI) applications like rehabilitation of stroke or spinal-cord injury patients, one of the prominent research topics is muscle movement detection from electroencephalogram (EEG). To endorse prosthetic organs there is a need to establish a strong relation between predetermined movements and BCI feedback, therefore, a robust technique for motor intention detection is required. The prosthetic muscles performance and efficiency depend critically upon the techniques being used to detect intended motion from EEG. Maximum efficiency for BCI applications can be achieved using detection techniques with limited latency and high accuracy. In this study, a robust motion detection technique using movement related cortical potentials (MRCPs) with improved accuracy is proposed. MRCP is extracted from multi-channel scalp EEG and it is a sluggish brain negative potential that is existing in EEG about 2 sec earlier than the execution and imagination of signaled and self-paced intended actions. Muscle movement can be detected by the earlier negative segment of MRCPs. In this technique after implementation of spatial filtering, short time Fourier transform (STFT) followed by threshold is used to detect movement intention using MRCPs from EEG. STFT provide us time frequency analysis of EEG signals and by it we detect a particular frequency associated with MRCPs and its corresponding time gives us information about the instant of muscle movement intention occurring. On the basis of statistical analysis, the proposed technique's results are better than the classic match filter technique. The True Positive Rate (TPR) of proposed algorithm is 75.81% and False Positive rate (FPR) is 45.46 % while TPR is 61.95% and FPR is 35.38% in the match filter technique. These statistical parameters exhibit the superiority of proposed algorithm. The results proved that the suggested algorithm detect the intention of motor movement more accurately. Due to its high performance it can be used in a closed-loop BCI systems for different applications.