Physiological Signal-Based Pain Recognition Using a Hybrid CNN–BiLSTM–GRU Architecture

Abstract

Automatic pain recognition using physiological signals plays a crucial role in modern healthcare systems. Traditional pain recognition methods rely heavily on handcrafted feature extraction, which requires medical expertise and limits scalability. To address these limitations, this paper presents an enhanced deep learning– based pain recognition framework that integrates feature extraction and classification into a unified architecture. The proposed extension incorporates a hybrid deep learning model combining Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and Gated Recurrent Unit (GRU) networks to effectively capture both spatial and temporal characteristics of physiological signals. Additionally, an ensemble stacking strategy is applied to strengthen prediction stability and overall performance. Multi-level contextual information is leveraged to provide a comprehensive representation of painrelated patterns, enabling accurate discrimination between pain and painlessness. Experimental results demonstrate that the proposed Extension CNN + BiLSTM + Bi-GRU model achieves superior performance, attaining an accuracy of 98.5%, thereby outperforming conventional machine learning methods and single deep learning models. The enhanced framework reduces dependence on extensive medical expertise, improves classification reliability, and offers a practical, scalable solution for automatic pain recognition in healthcare applications.