Heartfelt AI: Analyzing Emotions Through ECG Signals for Affective Computing

Affective computing is a branch of Artificial Intelligence (AI) that involves combining human emotions with AI technology. This helps in creating a human-computer interaction system by understanding human emotions, contributing significantly to human well-being. Through affective computing, it is also possible to conduct analyses of human mental health, which in turn can lead to enhancing human happiness and lifestyle. Emotion recognition systems play a pivotal role in enabling affective computing, with recent research focusing on utilizing psychophysiological signals to detect and differentiate emotions. This paper presents a comprehensive system that combines Electrocardiograph (ECG) signals, which reflect human heart activity influenced by the autonomous nervous system (ANS), with video stimuli to analyse and recognize emotions. The system utilizes the STM32 microcontroller and the AD8232 ECG sensor to capture and interpret ECG signals from a group of 30 participants. Additionally, it constructs a dataset of emotional states (normal, anxiety, and happiness) for these participants. Promising results are obtained, and the system further employs NanoEdge AI Studio to train a robust model. The trained model achieves an accuracy of 97.64% using a Random Forest classifier (RF). The model is then deployed on the STM32 microcontroller, integrated with an LED display, enabling real-time visualization and classification of emotions for new subjects with the help of ECG sensor.