Neural population models for perception of motion in depth

Area of Science:

• Neuroscience
• Computational Vision
• Perception Psychology

Background:

• Stereomotion perception relies on binocular cues.
• Changing disparity (CD) and interocular velocity difference (IOVD) are proposed mechanisms for motion-in-depth (MID).
• Understanding the neural basis of MID perception is crucial.

Purpose of the Study:

• To propose neurally plausible models for MID representation using CD and IOVD mechanisms.
• To investigate the distinguishability of CD and IOVD mechanisms.
• To explore the developmental learning of MID selectivity.

Main Methods:

• Developed computational models for MID velocity representation using population coding.
• Applied models to psychophysical stimuli (binocular image sequences, random dot stereograms).
• Simulated developmental learning of neural connectivity from natural binocular stimuli.

Main Results:

• CD and IOVD mechanisms are difficult to distinguish using standard psychophysical stimuli.
• Spatial connectivity in MID models can be learned from natural binocular input.
• Developmental models favor CD over IOVD for MID selective neuron emergence.

Conclusions:

• The study presents computational models for stereomotion perception via CD and IOVD.
• Findings challenge assumptions about distinguishing these mechanisms and highlight learning's role.
• Evidence suggests a developmental preference for the CD mechanism in MID processing.