Detecção do potencial relacionado à imaginação do movimento usando a filtragem de Kalman

Abstract

The brain-machine interface (BMI) is a communication system that uses brain signals to control the activation of external devices. The BMI is a two way interaction between people who have some motor impairment and the external environment. One way to improve the skills involved in developing BMIs is by means of the study of eventrelated potential (ERP) signals, which are obtained by recording the EEG signal. The identification of the ERP among the spontaneous electrical activity is a key step in implementing a BMI. The motor ERP, caused by a motor task, may be characterized by pre-event components, such as Contingent Negative Variation (CNV). Thus, the EEG signals recorded during mental tasks, such as voluntary movement execution (EEGTVM) or imagination of the movement (EEGIM), can, in principle, be used to drive external devices such as prostheses, orthoses or wheelchairs. In this work, EEG signals collected from 7 healthy individuals by means of 19 electrodes using the international 10-20 system were used. The contribution of this work is to estimate and detecting the ERP related to the imagination of the movement of the left hands index finger using an interactive bank with two Kalman filters (BIFK) in parallel. The first KF estimates the EEGTV signals, whereas the second one estimates the EEGTVM or EEGIM. This works methodology includes two steps: the modeling step and the estimation step. The BIFK reached classification rates of 65, 92, 89, 94, 90, 54, and 67 % for all individuals, respectively, for M=30 epochs. Initially, the EEG signal during the EEGIM contains M=30 epochs. Furthermore, we analyze the e_ect of the number of mental tasks repetitions on the detection of ERP. Froma practical standpoint, a smaller number of repetitions would make the BMI faster, since the individual would have to perform a smaller number of repetitions of the same task. Thus, in order to detect the signal ERP using fewer repetitions of EEGIM, this work proposes obtaining mathematical models from the averaging epochs of the EEG signal with M = 15, M = 10 and M = 5 during mental tasks. The results suggest that the BIFK is a promising method to detect the ERP related to the imagination of movement, making it a useful tool for the application of BMIs.