Related Experiment Video
Updated: Oct 4, 2025

Optrode Array for Simultaneous Optogenetic Modulation and Electrical Neural Recording
Published on: September 1, 2022
Mnemonic-opto-synaptic transistor for in-sensor vision system
Joon-Kyu Han1, Young-Woo Chung1,2, Jaeho Sim1
1School of Electrical Engineering, Korea Advanced Institute of Science and Technology, (KAIST) 291 Daehak-ro, Yuseong-gu, Daejeon, 34141, Republic of Korea.
A novel mnemonic-opto-synaptic transistor (MOST) integrates memory, photodetection, and synaptic functions for efficient in-sensor vision systems. This breakthrough enables faster, more energy-efficient machine vision by processing data directly within the sensor.
Area of Science:
- Materials Science
- Electrical Engineering
- Computer Science
Background:
- Conventional vision systems require separate image sensing and processing units, leading to bottlenecks in speed and energy efficiency.
- Artificial neural networks (ANNs) for machine vision demand significant computational resources and external memory.
- In-sensor processing offers a promising solution to overcome these limitations by integrating sensing and computation.
Purpose of the Study:
- To demonstrate a novel mnemonic-opto-synaptic transistor (MOST) with triple functions for in-sensor vision systems.
- To enable image sensing, memory storage, and synaptic weight updates within a single device.
- To advance machine vision capabilities by reducing reliance on external memory and processing units.
Main Methods:
- Fabrication of a vertical pillar-shaped transistor incorporating a charge trap layer (CTL) for memory functionality.
- Integration of photodetection, memory cell, and synaptic functions into a single MOST device.
- Demonstration of pattern recognition using fabricated MOSTs and software simulations for complex handwritten digits.
Main Results:
- The MOST successfully memorizes photoresponsivity as synaptic weights, enabling tunable light sensing.
- In-sensor processing was achieved, performing image sensing and signal processing simultaneously within the sensor.
- Enhanced speed and energy efficiency were observed compared to conventional vision systems due to on-sensor data preprocessing.
- Successful recognition of simple patterns and complex handwritten digits (MNIST) was demonstrated.
Conclusions:
- The developed MOST represents a significant advancement for in-sensor vision systems, integrating multiple crucial functions.
- This technology facilitates efficient machine vision by enabling on-sensor data processing, reducing the need for external memory.
- The MOST architecture holds potential for developing faster, more energy-efficient artificial intelligence hardware.
Related Concept Videos
Visual System
Once through the pupil, the light passes through the lens, a...
Photoreceptors and Visual Pathways
Bipolar Junction Transistor
Vision
MOS Capacitor
The metal gate is typically made from highly conductive materials such as aluminum or polysilicon. Beneath the metal gate lies a thin layer of...
Electrical Synapses
Gap junctions allow the current to pass directly from one cell to the next. In contrast, in the chemical synapse, the neurotransmitters carry the information through the synaptic cleft from one neuron to the next. They consist of two...

