DeepEmoNet: A Lightweight Context-Aware CNN for Multimodal Emotion Recognition
Engineering, Technology & Applied Science Research · 04.04.2026 · 16 (2) · 32844-32854
DOI: 10.48084/etasr.15258 ↗Kurz gefasst
Fachartikel in der Zeitschrift Engineering, Technology & Applied Science Research, online erschienen am 04.04.2026. Verfasst von Sumitra A. Jakhete und 1 weiteren Autor:innen. Die Arbeit nennt 29 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Multimodal emotion recognition in real-world environments remains challenging due to occlusions, class imbalance, and the high computational cost of existing deep models. This paper presents DeepEmoNet, a lightweight multimodal Convolutional Neural Network (CNN) designed to integrate facial, gait, scene, and socio-dynamic depth cues through an early-fusion architecture based on Depthwise Separable Convolutions (DSCs). The model aims to achieve robust emotion recognition while maintaining low computational overhead suitable for real-time applications. Experiments on the GroupWalk dataset comprising 3,544 annotated agents across 45 environments demonstrate that DeepEmoNet achieves 91.3% accuracy and 86.5% mean Average Precision (mAP), outperforming Inception V3, ResNet-50, MobileNetV2, and recent multimodal baselines. Extended ablation studies highlight the importance of contextual modalities and early fusion, with four DSC modules offering the best accuracy–efficiency balance. Inference analysis further shows a latency of 14.8 ms/frame (~67 frames per second (FPS)), supporting real-time deployment. Overall, DeepEmoNet offers an efficient, context-aware multimodal CNN framework for emotion recognition in surveillance, smart environments, and human–computer interaction.
Zitieren
APA
Jakhete, S. A., & Kulkarni, N. (2026). DeepEmoNet: A Lightweight Context-Aware CNN for Multimodal Emotion Recognition. Engineering, Technology & Applied Science Research, 16(2), 32844-32854. https://doi.org/10.48084/etasr.15258
Vancouver
Jakhete SA, Kulkarni N. DeepEmoNet: A Lightweight Context-Aware CNN for Multimodal Emotion Recognition. Engineering, Technology & Applied Science Research. 2026;16(2):32844-32854. doi:10.48084/etasr.15258
BibTeX
@article{Jakhete2026,
author = {Jakhete, Sumitra A. and Kulkarni, Nilima},
title = {DeepEmoNet: A Lightweight Context-Aware CNN for Multimodal Emotion Recognition},
journal = {Engineering, Technology & Applied Science Research},
year = {2026},
volume = {16},
number = {2},
pages = {32844--32854},
doi = {10.48084/etasr.15258}
}
EUVIMED