Article: Design of interactive feedback system for university online education classroom based on emotion recognition CNN model Journal: International Journal of Computational Systems Engineering (IJCSYSE) 2026 Vol.10 No.12 pp.1 - 12 Abstract: Virtual university classrooms present three interconnected problems that degrade learning quality: students struggle to assess their own emotional states, engagement levels remain low, and response latencies impair real-time interaction. It presents a multimodal framework for emotion detection and interactive response that combines image, audio, and text data with convolutional neural networks (CNN) to create a closed-loop system with both emotion estimation and the building of smart responses. The framework exhibits high accuracy in emotional detection, accompanied by low detection latency, with an average recognition latency of 149.2 milliseconds, making it suitable for real-time use. Instructors are notified, and students are offered personalised recommendations based on the emotional estimations, thus promoting personalised learning paths and improving interactions. Experimental results show an accuracy rate of 93.3% in emotional recognition, 3.01 times more student-to-student interaction within a 30-minute interval, 0.34 participation depth improvement, and 9.97% performance improvement. These results highlight the efficacy of enhancing real-time classroom interactions, stability, and learning assistance with the presented system, while also demonstrating a feasible solution for the improvement of smart classroom environments. Inderscience Publishers - linking academia, business and industry through research

Title: Design of interactive feedback system for university online education classroom based on emotion recognition CNN model

Authors: Hongyun Jie; Xiaozhe Li

Addresses: Hebei Open University, Shijiazhuang City, Hebei Province, 050080, China ' Hebei Open University, Shijiazhuang City, Hebei Province, 050080, China

Abstract: Virtual university classrooms present three interconnected problems that degrade learning quality: students struggle to assess their own emotional states, engagement levels remain low, and response latencies impair real-time interaction. It presents a multimodal framework for emotion detection and interactive response that combines image, audio, and text data with convolutional neural networks (CNN) to create a closed-loop system with both emotion estimation and the building of smart responses. The framework exhibits high accuracy in emotional detection, accompanied by low detection latency, with an average recognition latency of 149.2 milliseconds, making it suitable for real-time use. Instructors are notified, and students are offered personalised recommendations based on the emotional estimations, thus promoting personalised learning paths and improving interactions. Experimental results show an accuracy rate of 93.3% in emotional recognition, 3.01 times more student-to-student interaction within a 30-minute interval, 0.34 participation depth improvement, and 9.97% performance improvement. These results highlight the efficacy of enhancing real-time classroom interactions, stability, and learning assistance with the presented system, while also demonstrating a feasible solution for the improvement of smart classroom environments.

Keywords: online education; emotion recognition; convolutional neural networks; CNN model; interactive feedback; multimodal system; real-time optimisation; classroom interaction; student participation.

DOI: 10.1504/IJCSYSE.2026.157299

International Journal of Computational Systems Engineering, 2026 Vol.10 No.12, pp.1 - 12

Received: 14 Aug 2025
Accepted: 14 Sep 2026
Published online: 27 Oct 2026 *

Title: Design of interactive feedback system for university online education classroom based on emotion recognition CNN model

Keep up-to-date