A self-recognition and removal system of ocular artifacts based on FastICA

Ocular artifact (OA) is a common interference in electroencephalogram (EEG) processing. Due to the aliasing of the frequency bands of the two signals, the self-recognition and correction of OA are difficult. Therefore, this paper proposes a FastICA-K-means-CEEMDAN-WTD model based on fast independent component analysis and k-means cluster recognition artifacts, combined with full set empirical mode decomposition and wavelet threshold denoising, which can complete OA self-recognition correction for multi-channel EEG data. Compared with some OA denoising models based on Independent component analysis (ICA) , this model has the following advantages: 1. It can realize self-recognition without a lot of prior knowledge. 2. No pre-training is required. 3. It does not need to add additional ophthalmic electrical signals as a reference. 4. It can prevent the loss of EEG signals to the greatest extent. Based on the above method, the recognition and processing of the motion imagination EEG signal collected in the laboratory can obtain 94.79% accuracy.