基于机器学习的抑郁症特征提取与实现

Feature extraction and implementation of depression based on machine learning

文章将抑郁症脑电识别作为“云计算与数据挖掘”课程的实验内容,设计了利用机器学习进行抑郁症脑电识别诊断系统,利用信号处理方法进行脑电特征提取。时域采用基于统计特征的近似熵及非线性特征的模糊熵、频域采用基于脑电波段划分的功率谱密度进行特征提取,用以更加精准地提取抑郁症患者的脑电信号特征。最后利用机器学习方法,实现了对抑郁症的快速客观诊断。实验采用Python语言实现,实验结果表明近似熵特征取得了最佳分类结果。

This paper takes the EEG recognition of depression as the experimental content of the course of "Cloud computing and data mining", designs the EEG recognition and diagnosis system of depression by using machine learning, and extracts the EEG features by using the signal processing method. Approximate entropy based on statistical features and fuzzy entropy based on nonlinear features are used for feature extraction in time domain,and power spectral density based on EEG segment division is used for feature extraction in frequency domain, so as to extract EEG features of patients with depression more accurately. Finally, the machine learning method is used to realize the rapid and objective diagnosis of depression. The experiment is implemented in Python language.The experimental results show that the approximate entropy feature achieves the best classification results.