Discrimination Analysis and Movement Decision of the Prosthetic Arm by Using Monkey ECoGs Data Associated with Self-feeding Motions

概要

論文の詳細を見る
Brain Machine Interface (BMI) technology is a promising technology for the rehabilitation of patients with serious paralytic impairments. In particular, BMI using Electrocorticograms (ECoGs) provides an interesting new possibility because of its long-term stability, low degree of invasiveness, high spatial-temporal resolution and high signal per noise ratio (S/N ratio). Nevertheless, there are no BMIs with ECoG for devices that incorporate many degrees of freedom, such as the artificial hand. In this paper, as a preliminary experiment, we constructed a signal processing module and decision of movement of external devices such as a robot arm. We observed the self-feeding motions of a monkey (Macaca fuscata) employing ECoG electrodes. We applied the Fast Fourier Transform to the ECoGs and performed bandpass filtering to obtain amplitudes of a few frequency bands. We then implemented Linear Discriminant Analysis (LDA) to classify the motion patterns into four classes. Then, we decided the movement of the robot arm from the cumulative incidences of these classes. We have achieved two noteworthy results, namely a discrimination ratio between 53–61% and a delay of beginning times of 0.15 seconds.