Classification of EEG Signals on Standing, Walking and Running Dataset using LSTM-RNN

Undoubtedly one of the most important strands of the brain-computer interface (BCI) method is an alternate communication method via brain signals. BCI converts electroencephalogram (EEG) signals from a perception of activity in the brain into user action utilising software and hardware. BCI has piqued the interest of researchers in a wide range of disciplines, such as cognitive science, deep learning, pattern matching, drug treatment medicine, etc. Patients suffering from neuro and cognitive disorders can be assisted through BCI, potentially enabling communication via gestures or just mental imagination. In this paper, a novel combination of Discrete Wavelet Transform (DWT) for extracting the best features and Long Short-Term Memory (LSTM) based Recurrent Neural Network (RNN) is adopted for classifying the EEG signals acquired during standing, walking and running on a treadmill. The dataset used is freely downloaded from Open Science Framework repository. The proposed DWT-LSTMRNN method delivers 96.7% accuracy while classifying four different signals, and thus has the potential to be investigated further on BCI competition datasets that will pave way for a real-time application.