Depression level detection from facial emotion recognition using image processing

• Main Authors: N. A., Lili, M. R., Nurul Amiraa, M., MasRina, N., Nurul Amelina
• Format: Article
• Published: Springer Singapore 2022
• Online Access: http://psasir.upm.edu.my/id/eprint/100895/; https://link.springer.com/chapter/10.1007/978-981-16-8515-6_56
• Institution: Universiti Putra Malaysia

Adolescent depression is increasing daily at an alarming rate. Depression can be considered as a major cause of suicidal ideation and leads to significant impairment in daily life. Depression signs could be identified in peoples’ speech, facial expressions and in the use of language. We consider our study can help in the development of new solutions to deal with the early detection or diagnose of depression using facial emotion. Therefore, the objective of this study is to detect the level of depression using facial emotion via mobile application. This application provides user with the facial emotion recognition feature and a set of questions that are used to measure the level of depression of the user. This application will generate the total severity result of the depression alongside with the self-treatment and contact helplines recommendations. The added values of this application are the combination of facial emotion values and the questionnaire score to calculate the level of depression. The application would then recommend the types of treatment best suited for the user. We trained and tested classifiers to distinguish whether a user is depressed or not using features extracted from the user’s face expression. To predict depression, the face of the user will be captured, then using Gabor filters, the facial features are extracted. Classification of these facial features is done using Cascade and PCA classifier. The level of depression is identified by calculating the number of negative emotions present in the image captured. We used the F-measure scores as the performance score on the result gained. Overall result still in moderate level which is 76.7% due to low number of samples and features obtained.