Computational validation of the motor contribution to speech perception.

Area of Science:

• Neuroscience
• Cognitive Science
• Machine Learning

Background:

• Action perception and recognition are crucial for social interaction.
• The parieto-frontal network (mirror neuron system) links visual motion to motor representations.
• Machine learning for action identification improves with sensorimotor data.

Purpose of the Study:

• To review neurophysiological and machine learning studies on the motor system's role in speech perception.
• To demonstrate the benefit of motor data in automatic phone recognition.

Main Methods:

• Review of neurophysiological studies.
• Analysis of machine learning models trained with auditory and sensorimotor data.
• Comparison of automatic phone recognition performance.

Main Results:

• The motor system is specifically involved in speech perception.
• Automatic phone recognition accuracy significantly increases when motor data is included in classifier training.
• Models trained solely on auditory data perform less effectively.

Conclusions:

• The motor system plays a causal role in speech sound discrimination.
• Integrating sensorimotor information enhances speech perception and recognition technologies.
• Mirror neuron mechanisms are relevant for understanding speech processing.