"DETECTION OF DEPRESSION IN SPEECH PROJECT REPORT"

"Depression is a common mental disorder and one of the main causes of disability worldwide. Lackingobjective depressive disorder assessment methods is the key reason that many depressive patients can't betreated properly. Developments in affective sensing technology with a focus on acoustic features willpotentially bring a change due to depressed patient's slow, hesitating, monotonous voice as remarkablecharacteristics. Our motivation is to find out a speech feature set to detect, evaluate and even predictdepression. For these goals, we investigate a large sample of 300 subjects (100 depressed patients, 100healthy controls and 100 high-risk people) through comparative analysis and follow-up study. Forexamining the correlation between depression and speech, we extract features as many as possible accordingto previous research to create a large voice feature set. Then we employ some feature selection methods toeliminate irrelevant, redundant and noisy features to form a compact subset. To measure effectiveness of thisnew subset, we test it on our dataset with 300 subjects using several common classifiers and 10-fold cross-validation. Since we are collecting data currently, we have no result to report yet.In this project we are detecting depression from users post,user can upload post in the form of text file,imagefile, or audio file,this project an help peoplewho are in depression by sending motivated messages tothem.Now-a-days peoples are using online post service to interact with each other compare to human tohuman interaction.So by analysinguser post this application can detect depression and send motivationmessages to them.Administrator of this application will send motivation messages to all people who aredepression.To detect depression we are using SVM (support vector machine)algorithm which analyse userspost and give result as negative or positive.If users express depression words in post then SVM detect it as anegative post else positive post."

抑郁症是一种常见精神障碍，也是全球范围内导致残疾的主要诱因之一。目前缺乏客观的抑郁症评估手段，是诸多抑郁患者无法获得妥善诊疗的核心原因。聚焦声学特征的情感感知技术发展，有望借助抑郁患者语速迟缓、表达犹豫、语调单调的典型语音特征实现突破。本研究的核心动机是探寻可用于检测、评估乃至预测抑郁症的语音特征集。
为此，我们通过对比分析与追踪研究，对包含300名受试者的大样本队列开展调研：其中包括100名抑郁患者、100名健康对照者，以及100名高风险人群。为验证抑郁症与语音特征间的关联，我们基于既往研究尽可能多地提取特征，构建了大规模语音特征集；随后通过特征选择方法剔除无关、冗余与含噪特征，得到精简后的特征子集。为验证该新特征子集的有效性，我们基于包含300名受试者的数据集，采用多种常见分类器与10折交叉验证开展测试。由于目前数据采集工作仍在进行中，暂未获得可报告的研究结果。
本项目旨在通过用户发布的内容检测抑郁症，用户可上传文本文件、图像文件或音频格式的帖子。本项目可通过向抑郁人群发送励志信息，为其提供帮助。相较于线下人际互动，如今大众愈发依赖线上发帖服务开展社交互动。因此，通过分析用户发布的内容，本应用可实现抑郁症检测，并向存在抑郁倾向的用户推送励志信息。该应用的管理员可向所有被检测出抑郁的用户发送励志信息。本项目采用支持向量机（SVM, Support Vector Machine）算法检测抑郁症，该算法会对用户发布的内容进行分析，将其判定为阴性或阳性：若用户在帖子中使用了抑郁相关表述，则SVM将其判定为阴性（代表存在抑郁倾向），反之则判定为阳性。