Emotion Analysis Using Electrodermal Signals and Spiking Deep Belief Network.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Affective Computing

Background:

• Electrodermal Activity (EDA) is a physiological measure sensitive to emotional arousal.
• Distinguishing between arousal and valence dimensions in EDA signals is crucial for understanding emotional states.
• Existing methods may have limitations in accurately classifying complex emotional dimensions from EDA.

Purpose of the Study:

• To discriminate arousal and valence dimensions in Electrodermal Activity (EDA) signals.
• To evaluate the efficacy of a Spiking Deep Belief Network (SDBN) for emotion recognition using EDA.
• To explore the potential of SDBN in differentiating between autonomic and pathological conditions.

Main Methods:

• Utilized publicly available EDA datasets with varying arousal and valence dimensions.
• Preprocessed EDA signals through segmentation and channel normalization.
• Applied Spiking Deep Belief Network (SDBN) for feature extraction and classification of emotional states.
• Employed leave-one-out cross-validation to assess classification performance.

Main Results:

• The SDBN model demonstrated significant capability in discriminating between different emotional states.
• The network achieved superior classification performance for the arousal and valence dimensions of emotions.
• The proposed SDBN approach showed promise in differentiating autonomic and pathological conditions based on EDA patterns.

Conclusions:

• Spiking Deep Belief Networks are effective tools for analyzing Electrodermal Activity signals for emotion recognition.
• The SDBN approach offers a promising method for classifying emotional dimensions, particularly arousal and valence.
• This methodology holds potential for clinical applications in distinguishing physiological and pathological states.