The NEF-SPA Approach as a Framework for Developing a Neurobiologically Inspired Spiking Neural Network Model for...

Area of Science:

• Computational Neuroscience
• Neuroscience
• Artificial Intelligence

Background:

• Simulating speech production and perception with neurobiologically inspired models is challenging.
• Existing neural network models are limited, focusing on either cognitive-linguistic or sensorimotor aspects.
• Current models utilize second-generation, rate-based neuron approaches.

Purpose of the Study:

• To develop a third-generation spiking neuron neural network for comprehensive speech production modeling.
• To integrate cognitive and sensorimotor components within a unified neural model.
• To advance neurobiologically inspired computational models of speech.

Main Methods:

• Developed a neural model of speech production using the Neural Engineering Framework (NEF).
• Incorporated the Semantic Pointer Architecture (SPA) for constructing large-scale brain models.
• Utilized biologically grounded elements like spiking neurons, connections, and specialized modules.

Main Results:

• Constructed a fully functional model of speech production using biologically motivated spiking neuron micro-circuits.
• The model effectively simulates the entire speech production processing chain.
• Partially models speech perception and realistically simulates macroscopic speaking behavior.

Conclusions:

• The presented model offers a promising bottom-up approach to speech processing using micro-circuit neural network elements.
• It facilitates the generation of large-scale neural networks for complex cognitive functions.
• The model aligns with top-down design principles, correlating with functional imaging and electrophysiological data.