Assessment of Emotion-Based Depression Utilizing FERA-Net Model

• Bidyutlata Sahoo,
• Arpita Gupta,

Vol. 19, No. 10, October 31, 2025
10.3837/tiis.2025.10.007, Download Paper (Free):

• Facial Emotion Detection
• Depression Assessment
• video analysis
• Viola-Jones
• AlexNet
• FERA-Net
• Emotion classification
• Mental Health

Abstract

Depression detection and especially facial emotion recognition are crucial in psychological research and clinical evaluation. This paper focuses on summarizing a method of identifying six basic facial expressions in videos and its use in determining the degree of depression. Evaluation is done with the test set of samples from the Extended Cohn-Kanade (CK+) database containing different video sequences. First, the prompt enables the users to choose a video file containing facial expressions, and then the frames are analyzed. The Viola-Jones pre-trained cascade object detector is used to detect faces in each frame; and any detected faces are extracted as regions of interest. These ROIs are then passed through emotion categories by the Concatenated model which consists of LSTM & AlexNet (FERA-Net). When the positive emotion totals bear a ratio of less than P:N (where P represents the count of positive emotions such as happiness, and N represents the count of negative emotions such as sadness or anger), they imply severe depression or none; moderate to mild depression is implied when the positive-emotion totals bear a ratio of equal or greater than P:N. The user is constantly receiving information messages in the course of the process. The experimental results outperformed the normal results with an impressive of 98.9% for accuracy, 99% for precision, and 98.86% for F1 score. This method demonstrates the feasibility for accurate and automatic depression assessment with FER application.

Statistics

Cite this article