Recognition of affect based on gait patterns.

Area of Science:

• Human-computer interaction
• Biometrics
• Affective computing

Background:

• Recognizing human affect (emotional state) is crucial for advanced human-computer interaction.
• Gait, the manner of walking, is a potential non-intrusive indicator of affect.
• Current methods often require close proximity or direct interaction.

Purpose of the Study:

• To investigate the capability of gait to reveal a person's affective state from a distance.
• To compare different feature extraction techniques for affect recognition from gait.
• To assess person-dependent recognition accuracy for affective states.

Main Methods:

• Utilized motion capture data to analyze gait patterns.
• Employed Principal Component Analysis (PCA), Kernel PCA, Linear Discriminant Analysis, and General Discriminant Analysis for feature extraction and dimensionality reduction.
• Evaluated recognition accuracy for discrete affective states and affective dimensions (arousal, dominance).

Main Results:

• Person-dependent recognition of affect from gait achieved up to 95% accuracy using a single stride.
• Specific affective dimensions, such as arousal and dominance, were found to be reliably recognizable in gait.
• Feature extraction methods demonstrated effectiveness in reducing temporal gait information for classification.

Conclusions:

• Gait serves as a viable and accurate biometric modality for remote affect recognition.
• The findings support the integration of gait analysis into systems for monitoring and interaction.
• Potential applications include security, human-robot collaboration, and smart environments.