DETECTION OF MOVEMENT INTENTION FROM SINGLE-TRIAL MOVEMENT-RELATED CORTICAL POTENTIALS

Abstract

Stroke is the main cause of motor disabilities in the human beings regardless of their ages. For the rehabilitation of the stroke patients a method must be developed so that they can live an independent life. Previous studies have shown that by using different classifiers some methods are developed which can detect the movement intention of the patient. Previous studies on analyzing the detection of movement intention of different classifiers, such as Linear Discriminant Analysis (LDA), Support Vector Machines (SVM), employed detection based on decomposed data set. In thus study, we detected the movement intention using the entire data set. We used raw EEG signals of different subjects, and they were asked to imagine movement after some specified time. This movement generated the brain response, called movement-related brain potentials (MRCPs), which had a specific pattern. The recorded brain response was noisy and hidden within the EEG signal of the subjects. We developed a method in which we used matched filter combined with Neyman-Pearson detector to detect the movement intention. The matched filter was applied both in time and frequency domain. The True positive rate of 73% and 68% for local and global detector was achieved, respectively. This study addresses the problem of detecting the movement intention accurately, with limited latency. It was also concluded that matched filtering in time domain was faster as the impulse response of the filter was small.