An EEG signal classification method based on sparse auto-encoders and support vector machine

EEG signals, recording abnormal discharge of neurons in the brain, are widely used in epilepsy detection. In this paper, an EEG signal classification method based on sparse auto-encoders (SAE) and support vector machine (SVM) is proposed to greatly reduce the sample rate and enhance the efficiency of the vision detection. In practical application, sparse auto-encoder can get all the significant information at lower sample rate than sampled by Nyquist sampling principle. Due to this, it is widely used to extract higher layer features automatically. With the latter, it is used to obtain the high-dimensional pattern information of EEG signals, and map the input mode space into corresponding sparse space. This approach is precise enough to each sampling point rather than the conventional time window in the current researches and also has a better classification speed in comparison to other conventional methods. In order to ensure good classification rates (100%) for the EEG database, SVM is used to construct the generalized optimal classification hyper plane. Experimental result demonstrate that the classification rates in this work outperform the current state-of-the-art EEG seizure detection methods.