An EEG And Speech Based Psychological Stress Recognition Research

Psychological stress is one of the key factors to affect human health. When psychological stress one suffered is under excessive loads, psychological diseases like depression and anxiety disorder which are serious health hazards will be caused. As to the prevention of psychological diseases, accurately detection of psychological stress is of great significance. Psychological scales, as traditional methods to measure stress, are efficacious, but the disadvantages are appar-ent either. The self-assessment method is subjective, and the hormones measurement can not be widely promoted due to its invasion. With the development of electrophysiology and sensor technology, electroencephalography(EEG) and voice, as objective and non-invasive measure-ment, are favored by many researchers.This study comprises two parts, including EEG and voice. As to the EEG, we choose vol-unteers bear high stress as subjects. The frontal resting EEG is analyzed in order to explore the effectiveness of different combinations of features(linear and nonlinear features). Considering easily implement and fast calculation, the K nearest neighbor classifier and Naive Bayes classifi-er are utilized with a high classification accuracy. As to the voice, a mental arithmetic experi-mental task is designed to induce stress based on previous studies. We also extract voice acous-tics and prosody features, obtaining a high accuracy. Afterwards the linear discriminant analysis serves to select the optimal feature subset for classification. Additionally, a regression analysis is finished with the psychological scores and features for the research of stress level classification.The main contribution and innovation of this paper lies in:(1) It is found that nonlinear features of EEG such as Renyi entropy, LZ complexity, CO complexity, correlation dimension can significantly reflect the level of psychological stress of human body;(2) An experimental paradigm is improved based on mental arithmetic task, which is divid-ed into three levels of difficulty, in order to observe the trends of psychological stress;(3) It is discovered that psychological stress has a certain impact on the voice quality. The vibration frequency and jitter of vocal fold will increase in the task and decrease after the task, however, it can not restored to the original state in a short time. Additionally, the vibration fre- quency first increase and then decrease with the increase of complexity of the task.