Emotion Detection Using EEG and Voice Signals

This paper explores the feasibility of creating a polygraph capable of discerning emotional states in individuals by analyzing skin impedance, EEG waves, and voice changes in response to "YES" and "NO" questions. Utilizing a laboratory MP36 BIOPAC Acquisition Unit with high-resolution A/D sampling, this polygraph employs three channels: Galvanic Skin Response (GSR), EEG signals, and audio input via a microphone. Emotion recognition is based on a majority voting circuit that compares the responses from all three channels.The variation in the amplitude of the skin conductivity is detected by comparing the derivative of the signal with a threshold value, and for the classification of EEG and vocal signals we use feedforward neural networks. To reduce the neurons in the input layers of the networks, the signals are processed using the Discrete Cosine Transform (DCT). Our findings reveal promising results in emotion detection via these multiple channels, offering potential applications in fields like human-computer interaction and emotional state monitoring.