Depression signal correlation identification from different EEG channels based on CNN feature extraction

•This paper proposes a convolutional neural network (CNN) based correlation recognition method for depression signals.•The method uses labeled multi-channel Eeg signals as data.•Eeg signals of each channel were divided into neural network training data sets, and AlexNet network was used to train these data. Then, the depression channels were classified according to the training samples.•The accuracy function and loss function are used to evaluate the classification index. Depression is a mental illness and can even lead to suicide if not be diagnosed and treated. Electroencephalograph (EEG) is used to diagnose depression and it is more complexity to extract the features from all the multimodal channel information . In order to simplify the diagnose process and detect clinical depression, the EEG channels with strong depression information should be identified firstly. Therefore, a depression signal correlation identification method based on convolutional neural network (CNN) is proposed. In the method, the labeled multi-channel EEG is used as data. The EEG signals of each channel are divided into neural network training data set and these data is trained by AlexNet network. Then the correlation classification of each channel for depression is identified based on the trained sample. Accuracy and loss functions are used to evaluate classification index.Conversely, the correlation is lower. An experiments is conducted and the results show that the correlation is not consistent. A few of channels are strongly correlated with depression, such as 13, 17, 28, 40, 46, 66 and 69. These EEG channels are selected to diagnose depression.