Severe Major Depression Disorders Detection Using AdaBoost-Collaborative Representation Classification Method

Depression has become one of the most common psychological diseases, speech-based depression detection and severity recognition is a new trend in recent years. However, data insufficient issue and the significant imbalance among the different classes are the main challenge in the area at present. To...

Full description

Saved in:
Bibliographic Details
Published in:2018 International Conference on Sensing,Diagnostics, Prognostics, and Control (SDPC) pp. 584 - 588
Main Authors: Jingwen Zhang, Haochen Yin, Jinfang Wang, Shuxin Luan, Chang Liu
Format: Conference Proceeding
Language:English
Published: IEEE 01.08.2018
Subjects:
Acoustics
adaboost algorithm
class imbalance
Classification algorithms
Collaboration
collaborative representation
Dictionaries
Indexes
major depression disorders
Sensitivity
small sample size
Training
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Depression has become one of the most common psychological diseases, speech-based depression detection and severity recognition is a new trend in recent years. However, data insufficient issue and the significant imbalance among the different classes are the main challenge in the area at present. To solve the problem of sample insufficient and class imbalanced, this paper combines AdaBoost and collaborative representation (AdaBoost-CRC) to detect Severe Major Depression Disorders (SMDD). Firstly, Mel-Frequency Cepstral Coefficient (MFCCs) were extracted from the subject's speech; Then, aiming at the data imbalance issue, AdaBoost-CRC classifier structure was created in which AdaBoost was used to discriminate the result of each weak classifier according to its weight. The experimental framework of leave-one-speaker-out cross validation was adopted to assess the method's performance. The evaluation data comes from the Audio/Emotion Challenge and Workshop (AVEC) 2013 dataset. Experimental results show that the accuracy and sensitivity of AdaBoost-CRC are better than CRC's.
DOI:10.1109/SDPC.2018.8665013