Single-channel EEG-based prosthetic hand grasp control for amputee subjects

This article explores the use of single-channel single-trial EEG signals for natural control of prosthetic hand grasp. It is natural in the sense of that the desired movement is what the subject intends to do. The motor tasks to be intended are the imagination of hand grasping and opening. For prosthetic hand grasp control, the discrimination of the resting state and the imagined voluntary movement is important that has been disregarded in the research area of BCI. This work provides a design for discriminating the resting state and the motor task imagery. To date most researchers have designed and test BCI system on normal subjects. In this work, the experiments were conducted on people with severe physical disabilities. One of the major problem in developing real-time BCI is the eye blink suppression. In this work, the eye blink artifact is suppressed by a neural adaptive noise canceller. This is a concern in real time application. We employ the multilayer perceptron with back-propagation learning rule for EEG classification. Preliminary results indicate that the classification accuracy of the EEG patterns at primary motor cortex and occipito-temporal recording sites is higher than that at other sites. An average correct classification rate of 83% is achieved using samples of the single-channel EEG signal.