Motion-based prediction explains the role of tracking in motion extrapolation

Area of Science:

• Neuroscience
• Computational Vision
• Perception

Background:

• The visual system typically processes a continuous stream of sensory information.
• Intermittent sensory input, such as during blinks, poses a challenge to maintaining a coherent perception of motion.
• Existing evidence suggests neural mechanisms for motion extrapolation exist to handle fragmented visual data.

Purpose of the Study:

• To model how the visual system extrapolates object trajectories during sensory blanks using motion-based prediction.
• To investigate the role of motion coherence priors in integrating information during stimulus absence.
• To compare model simulations with experimental behavioral and neural data on motion extrapolation.

Main Methods:

• Developed a computational model simulating motion-based prediction for trajectory extrapolation during visual blanks.
• Simulated tracking velocity responses to assess the model's predictive capabilities.
• Tested the model's response to noisy stimuli to analyze the influence of sensory noise on tracking behavior.

Main Results:

• The model demonstrated that motion-based prediction allows the visual system to recover trajectory information after a sensory blank.
• Motion-based prediction functions as a global gain control mechanism, enabling a switch between smooth tracking and binary tracking (tracked or lost).
• Tracking behavior emerges only after sufficient trajectory information is accumulated and remains robust to blanks and noise once established.

Conclusions:

• A local prior implementing motion-based prediction is sufficient to explain global neural and behavioral observations in motion extrapolation.
• Tracking is essential for motion extrapolation, and its performance deteriorates beyond a critical sensory noise level.
• Further research is needed to explore the specific role of noise in the mechanisms of motion extrapolation.