Bio-inspired cross-modal super-additive plasticity for seamless visual processing-in-sensory and -in-memory.

Area of Science:

• Materials Science
• Neuroscience
• Computer Engineering

Background:

• Conventional optoelectronics lack super-additive integration for bio-inspired visual perception.
• Edge intelligence requires efficient hardware for processing-in-sensory and -in-memory applications.

Purpose of the Study:

• To design a novel hardware platform for bio-inspired cross-modal visual perception.
• To overcome limitations of linear function combinations in existing hardware.

Main Methods:

• Utilized molybdenum disulfide channel devices for processing-in-sensory and -in-memory.
• Employed electric field-assisted photogenerated carrier tunneling in a floating gate photoelectric device array.
• Integrated a non-volatile four-transistor ternary content-addressable memory circuit array for pattern recognition.

Main Results:

• Demonstrated cross-modal correlation photoelectric signal processing with super-additive behavior up to 10^3 times.
• Achieved significant time-dependent plasticity for visual encoding and perception enhancement.
• The memory cell maintained a resistance ratio of 10^5 and lookup durability of 10^12.

Conclusions:

• The developed hardware platform enables seamless visual processing-in-sensory and -in-memory.
• This technology holds potential for ubiquitous visual edge intelligent systems.
• The bio-inspired approach enhances visual perception and pattern recognition capabilities.