Identifying motor imagery in the act of fisting using Poincaré cut for use in brain computer interface

The main idea for this paper is to help disabled people who need to use prosthetic hands. In this paper, we have used EEG signals from the ۱۰۷ healthy volunteer, who were asked to do some tasks such as right/ left fist movement, imagination of fist movement and relaxed state with closed eyes (as a base situation). We have applied Linear Discriminant Analysis (LDA) and Non-Linear Analysis particularly Poincare analysis to extract features. We evaluate our results using Kruskal Wallis test and measuring the p-value of each parameter. Good features with approximate zero p-value result have recognized. The most beneficial features was Poincare analysis parameters such as sequence of EEG points under and above of 𝑥=𝑦 line in Poincare plot. Then we have applied neural network to classify our three groups. The average of classification accuracy was 𝟗𝟎.𝟔% for imagination and real hand movement classification, and absolute accuracy of ۱۰۰% for base and imagination state classification.

Keywords:

EEG signal processing , Poincare , Imagination of hand movement , Hand prosthetics and rehabilitation , BCI , NNs