Joints and their relations as critical features in action discrimination: evidence from a classification image method

Area of Science:

• Human visual perception
• Biomechanics
• Machine learning in neuroscience

Background:

• Action classification, like distinguishing walking from running, is complex due to numerous biological motion features.
• Previous hypothesis-driven research struggled to isolate key discriminating features.

Purpose of the Study:

• To develop a hypothesis-free method for identifying features used in bipedal action discrimination.
• To investigate how viewing conditions affect the selection of these features.

Main Methods:

• Utilized a classification image method with ambiguous walker-runner stimuli.
• Analyzed approximately 1,000 trials per subject to identify key movement features.
• Tested feature importance under inverted and depth-rotated viewing conditions.

Main Results:

• Identified three critical feature types: critical joint, structural, and relational.
• Critical joint features (feet, wrists) were consistently important across participants.
• Feature importance varied with viewing conditions; critical joint features dominated inverted views, while critical joint and relational features were key for depth-rotated views.

Conclusions:

• The human visual system flexibly uses a combination of joint, structural, and relational features for action discrimination.
• The selection of these features is adaptable based on environmental context and viewing parameters.
• This hypothesis-free approach reveals novel insights into biological motion perception.