Hybrid architectures based on Quantum Wavelet Neural Network for Emotion Recognition as a tool to support Personalized Therapies
Arianne Sarmento Torcate, Maíra Araújo de Santana, Flávio Secco Fonseca +5 more
Human emotions play a fundamental role in social interaction and effective communication between individuals. In the healthcare context, tools based on emotion recognition are promising for therapeutic applications, especially for patients with difficulties in expressing their emotions, who are often unable to communicate the discomfort associated with treatments. This work presents a comparison between hybrid architectures composed of a Quantum Wavelet Neural Network (QWNN) and a Random Forest, using the FER2013 database for training and testing. Our best configuration, with a four-layer QWNN and Random Forest, achieved 70.77% accuracy in the training phase and 81.21% in the testing phase, even using only 34.06% of the features extracted by the QWNN (436 out of 1.280). In addition to its high performance in feature extraction, the QWNN has great potential for synergy with quantum computing due to its ability to represent information at multiple scales and frequencies simultaneously. Finally, the proposed architecture aims to integrate an emotion recognition system based on facial expressions in elderly people (with and without dementia) for personalizing therapies, paving the way for future research that explores Wavelet-based neural networks and Quantum Computing in the health area.
Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC)•Jul 14, 2025•0 citations