Balancing prior knowledge and sensory data in a predictive coding model of coherent motion detection.

Area of Science:

• Computational neuroscience
• Visual perception

Background:

• The visual system, particularly MT neurons, uses surround suppression for global motion detection.
• Predictive coding principles are fundamental to understanding brain function.

Purpose of the Study:

• To introduce a neurobiologically inspired computational model of coherent motion detection.
• To investigate the role of prior knowledge and sensory data in motion perception.
• To explore the model's implications for understanding visual processing deficits.

Main Methods:

• Developed a computational model based on the predictive coding algorithm.
• Simulated motion structure decomposition into individual and shared sources.
• Tested the model using random dot stimuli under varying noise levels.

Main Results:

• The model demonstrates the balance between sensory input and prior knowledge in motion detection.
• Increased noise levels prolonged motion estimate stabilization time, mirroring psychophysical findings.
• Excessive prior knowledge hindered stabilization, while optimal integration improved accuracy and efficiency.

Conclusions:

• The model provides insights into the neural mechanisms of coherent motion detection.
• Findings suggest that prior knowledge plays a crucial role in modulating motion perception.
• The model may help explain motion detection impairments in conditions like schizophrenia.