EUVIMEDEUVIMED entdecken
Crossref

An Intelligent Deep Learning Model for Human Emotion Recognition via Contextual Cues

Tejas Sambhaji Nalawade, Monika D. Rokade, Sunil S. Khatal, Puja S. Gholap

International Journal of Drug Delivery Technology · 06.06.2026 · 16 (47s)

DOI: 10.25258/ijddt.16.47s.40 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Drug Delivery Technology, erschienen am 06.06.2026. Verfasst von Tejas Sambhaji Nalawade und 3 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Automated emotion recognition from multimodal human signals has emerged as a critical enabler of contextaware intelligent systems, spanning domains from clinical mental health monitoring to adaptive human-computer interfaces. Conventional unimodal approaches suffer from inherent ambiguity when confronted with real-world variability such as partial facial occlusion, pose changes, and individual expressive style differences. The present work introduces a multi-modal deep learning architecture that concurrently processes facial, linguistic, and physiological contextual streams to achieve more reliable emotion inference. A ResNet-50 convolutional network handles spatial feature learning from facial imagery, a two-directional LSTM encoder captures the sequential dynamics of facial expressions over time, and a BERT transformer encoder captures semantic contextual information from speech transcriptions. A learnable cross-attention fusion mechanism subsequently integrates these heterogeneous representations prior to a seven-class emotion classifier. Comprehensive evaluation on the FER-2013, RAVDESS, and DEAP benchmark corpora demonstrates an overall accuracy of 91.7%, exceeding the best competing single-modality system by 7.2 percentage points. Ablation studies isolate the contribution of each processing stream and confirm that the cross-attention module provides a 2.6 percentage point improvement beyond naive feature concatenation. Performance under Gaussian noise corruption at signal-to-noise ratios down to 10 dB further validates the noise tolerance advantages of multi-modal integration.

Zitieren

APA

Nalawade, T. S., Rokade, M. D., Khatal, S. S., & Gholap, P. S. (2026). An Intelligent Deep Learning Model for Human Emotion Recognition via Contextual Cues. International Journal of Drug Delivery Technology, 16(47s). https://doi.org/10.25258/ijddt.16.47s.40

Vancouver

Nalawade TS, Rokade MD, Khatal SS, Gholap PS. An Intelligent Deep Learning Model for Human Emotion Recognition via Contextual Cues. International Journal of Drug Delivery Technology. 2026;16(47s). doi:10.25258/ijddt.16.47s.40

BibTeX
@article{Nalawade2026,
  author = {Nalawade, Tejas Sambhaji and Rokade, Monika D. and Khatal, Sunil S. and Gholap, Puja S.},
  title = {An Intelligent Deep Learning Model for Human Emotion Recognition via Contextual Cues},
  journal = {International Journal of Drug Delivery Technology},
  year = {2026},
  volume = {16},
  number = {47s},
  doi = {10.25258/ijddt.16.47s.40}
}
Kostenlos starten →