【作者】 赵国庆；

【导师】 胡斌；

【作者基本信息】 兰州大学 ， 计算机软件与理论， 2013， 硕士

【摘要】 社会经济的飞速发展使得人们面临的工作、生活和性格中的压力剧增。如何在进一步发展为精神紊乱病症之前,发现压力并及时采取措施是十分必要的。例如,压力的监测在检测和干预抑郁症过程中起到了关键的作用。传统的压力监测手段是通过各种用户自评量表,但它的突出问题是难以避免的主观性。同时,缺乏一套可应用于日常生活中的压力监测系统。为了能够在日常应用中准确客观的对压力进行监测,本文开发了一套在线压力监测系统。本系统通过采集Fpz一个电极的脑电信号,能够对用户的压力状态进行客观的评价并进行长期在线监测,同时可辅助医师对用户状态进行评估。系统采用c／s框架,基于soap协议,客户端分为用户界面和医师界面,用户界面包括信号采集、量表填写、历史记录查看、医师交互模块。医师界面包括查看用户历史记录和与用户交互模块。服务器端主要负责对用户数据进行有效管理,以及脑电信号的处理。本文通过实验,选择出系统中使用的有效的脑电特征和算法。首先采集被试脑电信号,进行去噪、特征提取和分类,筛选出三个对压力分类有效地脑电信号特征(分别是LZ复杂度,alpha相对功率,alpha相对功率/beta相对功率),并通过结果对比最终确定K最近邻分类器为系统中的分类算法。为了直观的显示压力水平,引入了压力指数这一概念。更多还原

【Abstract】 The rapid development of social economy makes people facing increasing stress from work, life and character. Before further development to mental disorder illness, how to detect the stress and take timely measures are very necessary. For example, the monitoring of the stress plays a key role in the process of detection and intervention of depressionThe traditional mean of stress monitoring is a variety of user self-rating scale, but its outstanding problems is the difficulty of avoiding subjectivity. Meanwhile, we need an online stress monitoring system that can be applied in daily life.In order to monitor stress easily and objectively in daily life, we build a pervasive online stress detection system. By collecting EEG from Fpz point, the system can make objective evaluation and conduct long-line monitoring of users’ stress, at same time, can assist physicians to assess the users’ state. Based soap protocol, the proposed system adopts client/server framework. Client-side consists of "User Panel" and "Doctor Panel". The "User Panel" integrates modules including biological signals acquisition, self-assessment questionnaire, history record and doctor chatting. The "Doctor Panel" includes history record module and user chatting panel. Server-side is mainly responsible for the effective management of users’data, as well as EEG processing.For the purpose of adopting effective EEG features and algorithms in the system, an experiment has been conducted. After collecting subjects’ EEG signal, denoising, feature extraction and classification, three features effective for stress classification were screened out, namely LZ-complexity, alpha relative power and the ratio of alpha power to beta power. By comparing results the k-nearest neighbor classifier were determined as classification algorithm in system. Meanwhile, we introduced the stress index for indicating stress level intuitionisticly.更多还原