PAIN RECOGNITION WITH PHYSIOLOGICAL SIGNALS USING MULTI-LEVEL CONTEXT INFORMATION

Abstract

The project aims to develop an automatic pain recognition system in healthcare, removing the dependence on medical expertise for manual feature extraction from physiological signals. This shift addresses the limitations of conventional methods, making pain recognition more accessible and widely applicable. The proposed solution introduces a deep learning model that uniquely combines the roles of feature extraction and classification. By leveraging the strengths of deep neural networks, this approach streamlines the process, eliminating the need for separate feature engineering and classification steps commonly found in traditional methods. The project introduces a novel aspect by incorporating multi-level context information for each physiological signal. Unlike uni-level context information used in prior approaches, this multi-level understanding aims to provide a more nuanced perspective on pain and painlessness. It enhances the discriminative power of the model by considering various levels of context within the physiological signals. The deep learning approach demonstrated in the project showcases its superiority in handling physiological signals for pain recognition. By eliminating the need for explicit feature engineering by medical experts, the model can autonomously learn and extract relevant features directly from the data. This not only marks a significant advancement over conventional methods but also enhances the efficiency and accuracy of pain recognition based on physiological signals. The project's include a “Stacking Classifier” and hybrid model “CNN+BILSTM+GRU”, in which stacking classifier got 99% accuracy for enhanced Pain Recognition .