Abstract:
This research provides the design and development of a portable, lowcost Electroencephalography (EEG) based real-time emotion recognition system. Emotion recognition is crucial for comprehending human behavior, psychological monitoring, and Human-Computer interaction (HCI). Compared to traditional emotion recognition methods such as questionnaire and manual observation, EEG provides a convenient and objective technique that directly measures brain activity. In this study, a Convolutional Neural Network (CNN) model is applied to differentiate emotional states among positive, neutral, and negative classes by learning spatio-temporal features automatically from the EEG signals. A hardware system was built by combining a 24-channel EEG headset with a microcontroller of ESP32 and an ADC of ADS1256. Signal acquisition and transmission can then be effectively realized. In order to obtain better signal quality, the signals are frst pre-processed with bandpass fltering (0.5-50 Hz), artifact rejection and normalization, followed by feature learning and classifcation of the CNN model. The experimental results show that the developed system achieves stable and reliable accuracy in real-time and quickly response. The system can effectively recognize emotion states like happiness and sadness and can function outside a lab environment. Generally, this developed technique represents a practical and low-cost solution for real world applications such as mental status supervision, stress detector, and wearable Human-Computer Interface system.
