Multi-Class Support Vector Machines for Brain Neural Signals Recognition

It is promising to control neuroprosthetic devices by the activity of cortical neurons when appropriate algorithms are use to decode intended movement. In this paper, a multi-class support vector machines (SVMs) algorithm of a binary tree recognition strategy is used to analyze the motor cortical neuronal signals. The neural ensemble data were recorded simultaneously with kinematics of arm movement while the monkey performed reaching tasks from the center position to eight peripheral targets in a three-dimensional (3D) virtual environment. The SVMs based method was applied to classify the neural ensemble firing rate patterns into eight classes. The performance of the SVMs based neural activity recognition was compared with that of the learning vector quantization (LVQ) approach. The results show that the SVMs can achieve higher accuracy with less computational time, which demonstrates that the SVMs algorithm is a suitable approach for brain neural signals recognition