Read the full-text of this article for free

Title: Construction of mental health analysis model based on multi-modal feature learning and fusion network

Authors: Sujing Li; Suya Liu; Maochun Wu

Addresses: Department of Medical Care and Health, Jining Polytechnic, Jining, 272000, China ' Department of Cultural Studies and Public Administration, Jining Polytechnic, Jining, 272000, China ' Department of Cultural Studies and Public Administration, Jining Polytechnic, Jining, 272000, China

Abstract: This paper presents a mental health analysis model using a multi-modal feature learning and fusion network to improve assessment accuracy. It integrates data from text, images, and speech, processed with CNNs, RNNs, and LSTMs for feature extraction and fusion. Experimental results show the multi-modal model achieves 85% classification accuracy, outperforming single-modal models (75%). Analysis of feature weights indicates audio and visual modalities significantly influence emotional fluctuation (30%) and coping ability (40%), while physiological signals are crucial across all traits. The model enhances assessment comprehensiveness and offers effective support for early diagnosis and personalised intervention.

Keywords: multimodal feature learning; deep learning; mental health analysis; feature fusion; privacy protection.

DOI: 10.1504/IJICT.2026.151685

International Journal of Information and Communication Technology, 2026 Vol.27 No.10, pp.1 - 21

Received: 26 Aug 2025
Accepted: 16 Oct 2025
Published online: 13 Feb 2026 *