by Purdue University researchers are developing an artificial retina that mimics the human visual system in order to improve the efficiency of computer vision. The device, known as an organic electrochemical photonic synapse, is designed to sense change and is more energy-computationally efficient than traditional digital camera systems used in applications such as self-driving cars and autonomous robots.
Computer vision systems require large amounts of energy, creating a bottleneck for widespread use. With the goal of using biomimicry to address the challenge of dynamic imaging with less data processing, the researchers drew inspiration from the light perception in retinal cells. They built a prototype device that triggers an electrochemical reaction when exposed to light, gradually strengthening and creating a memory of the light information received.
This memory can potentially reduce the amount of data that needs to be processed to understand a moving scene, making the approach more energy-efficient and error-tolerant compared to conventional computer vision methods. The researchers believe their device more closely mimics how the human visual system works and has greater potential as the foundation of a device for human-machine interfaces.
In a normal computer vision system, data must be transferred from memory to processing and back, resulting in significant time and energy consumption. The organic electrochemical photonic synapse incorporates light perception, light-to-electric signal transformation, and on-site memory and data processing, eliminating the need for data transfer.
Currently, digital cameras are the foundation of computer vision for robotic and autonomous devices. These cameras use photosites made of crystal silicon to absorb photons and convert light into an electrical signal. However, this approach requires analyzing all available light information, regardless of whether the scene changes or remains static.
In contrast, the artificial retina developed by the Purdue researchers, inspired by human vision, has relatively low resolution but is well-suited for sensing movement. The prototype device has a resolution of a few hundred microns, and the researchers believe it could be improved to around 10 microns.
Rather than directly converting light to an electrical signal, the researchers convert light to a flow of charged atoms called ions. This mechanism is similar to the way retinal cells transmit light inputs to the brain. The presence of light attracts positively charged ions, creating a charge imbalance. With repeated exposure to light, this charge imbalance increases, distinguishing between static and dynamic scenes. When the light is removed, the ions remain in their charged configuration for a short period of time, allowing for motion sensing and memory capabilities.
The performance of the researchers’ electrochemical transistor is superior to other optoelectronic devices with integrated light perception and memory capabilities. The charge imbalance in their device increases smoothly and steadily with repeated exposure to light and decays more slowly. The researchers plan to use flexible materials in future iterations, potentially creating a wearable and bio-compatible version of the device.
Note:
1. Source: Coherent Market Insights, Public sources, Desk research
2. We have leveraged AI tools to mine information and compile it
