Abstract
In recent years, Brain–computer interfaces (BCIs) have been widely studied and become popular research topics on the applications of many fields. Several electroencephalography (EEG) features have been discovered so far, such as event related synchronization/desynchronization (ERS/D). By recording the EEG signal from BCI and processing based on ERS/D, the objective to analyze and estimate human motor intention is accessed. However, EEG signal is unstable and non-linear, and has a low tolerance for noise. Therefore, how to process the signal and extract features properly is the main topic in this field. In this study, we propose a classification algorithm for L/R hand motor imagery (MI) based on ERS/D. Several techniques are applied, including Short-Time Fourier Transform (STFT), cosine normalization, feature extraction based on ERS/D, Linear discriminant analysis (LDA), k-nearest neighbors algorithm (kNN), Support Vector Machine (SVM), Artificial Neural Network (ANN) and Leave-One-Out Cross-Validation (LOOCV). We mainly focus on designing the methods to extract the features, which are novel and robust, and propose how to optimize the parameters in it. Results show that our proposed method has a better performance than others with same database, and the classification accuracy 97.62% for L/R hand MI is achieved.