Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{207335,
        author = {Prajwal K and K. Deepak and Samruddhi S Hegade and Thanu Shree H M and Dr. Arjun U},
        title = {ResolveNet: A Multimodal Geometric Fusion Framework for Affective Masking Resolution in Real-Time Engagement Monitoring},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {3},
        pages = {703-713},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=207335},
        abstract = {Conventional engagement monitoring systems in Human-Computer Interaction (HCI) predominantly rely on single-modality analysis—typically facial expression recognition—and are therefore inherently susceptible to a phenomenon termed "affective masking," wherein a user's visible facial expression consciously or unconsciously contradicts their true internal emotional state. This misalignment generates false-positive engagement metrics that critically reduce system reliability in real-world deployments. This paper introduces ResolveNet, a late-fusion multimodal architecture that resolves affective ambiguity through a geometric decision layer. Visual features extracted by a Convolutional Neural Network (CNN) and textual sentiment scores derived from lexicon-based analysis are co-embedded into a shared 14-dimensional feature vector. A K-Nearest Neighbors (KNN) classifier, trained on curated conflict-resolution seed samples, serves as a non-parametric geometric boundary that synthesizes conflicting modal signals into a coherent engagement label. Real-time performance at 30 frames per second on standard CP sfsdf
hardware is achieved through Haar Cascade face localization, grayscale feature extraction, and deque-based temporal mode-voting. Experimental evaluation demonstrates that ResolveNet achieves an accuracy of 89.4% on multimodal conflict scenarios, outperforming unimodal baselines by 14.2 percentage points, and correctly resolves affective masking events with a precision of 91.7%. Our framework elevates engagement detection from a simple classification task to a Signal-Ambiguity Resolution problem, with direct applicability to online education, telehealth, and adaptive HCI systems. The proposed system further demonstrates computational efficiency that enables deployment on edge devices without hmmspecialized hardware acceleration, broadening its accessibility across educational and healthcare infrastructures in resource-constrained environments.},
        keywords = {Affective computing, affective masking, engagement monitoring, KNN fusion, late-fusion architecture, multimodal emotion recognition, real-time HCI, conflict resolution, human-computer interaction.},
        month = {August},
        }