Abstract: Schizophrenia is a common serious mental disease, which has seriously affected the quality of human life for many years. Therefore, accurate diagnosis is a prerequisite for the treatment of the disease. To solve the problem, this paper proposes a classification method of schizophrenia electroencephalogram（EEG） signals based on brain complexity and deep learning, aiming at discovering hidden distributed features in data. Contrary to the standard EEG data analysis technology that ignores spatial information, the time series of EEG signals are firstly divided into different frequencies, and fuzzy entropy（FuzzyEn） is used to extract the complexity features of each frequency band, then the feature vectors are constructed according to the spatial position of the electrode and input into the Convolution Neural Network（CNN） to train the classification model, so as to automatically identify whether the subject is diseased. The results show that the experimental method is effective, and the classification accuracy reaches 99.16%.

Key words: schizophrenia, fuzzy entropy, Convolution Neural Network（CNN）, classification

摘要： 精神分裂症是一种常见的重性精神疾病，多年来严重影响人类的生活质量，因此，对该病的准确诊断是治疗疾病的前提。针对以上问题，提出一种基于大脑复杂性和深度学习的精神分裂症脑电信号（EEG）分类方法，旨在发现隐藏在数据中的分布式特征。与忽略空间信息的标准脑电数据分析技术相反，首先将脑电信号的时间序列进行分频处理，并将每个频段的时间序列用模糊熵（FuzzyEn）进行特征提取，按照电极的空间位置构成特征向量，并将特征向量输入到卷积神经网络（CNN）中训练分类模型，自动识别受试者是否患病。实验结果表明，基于模糊熵和卷积神经网络的分类方法是有效的，分类准确率达到了99.16%。

关键词: 精神分裂症, 模糊熵, 卷积神经网络, 分类