The Framework and Memristive Circuit Design for Multisensory Mutual Associative Memory Networks

Area of Science:

• Neuroscience
• Computer Engineering
• Materials Science

Background:

• Biological brains exhibit remarkable associative memory capabilities, integrating multisensory information.
• Existing artificial systems often struggle with bidirectional, multimodal associative memory.
• Memristive devices offer potential for hardware implementation of neural functions.

Purpose of the Study:

• To propose a multisensory mutual associative memory network framework.
• To design a memristive circuit mimicking biological associative memory.
• To achieve bidirectional association across multiple sensory modalities.

Main Methods:

• Developed a three-module circuit: storage neurons, memristive synapses, and retrieval neurons.
• Encoded multimodal information into spike firing rates.
• Utilized nonvolatile memristors for synaptic weight adjustment and learning.
• Simulated associative memory properties like acquisition, extinction, and consolidation.

Main Results:

• The circuit successfully demonstrated bidirectional, multimodal associative memory.
• Cross-modal association between image and audio recognition was achieved.
• Intramodal association for episodic memory (multiple images) was simulated.
• The circuit was validated as hardware-friendly with power and area analysis.

Conclusions:

• The proposed framework and memristive circuit effectively mimic brain-like associative memory.
• This approach enables bidirectional, multimodal information association.
• The hardware-friendly design supports potential applications in neuromorphic systems and intelligent robotics.