(C) Inside Telecom

BEIJING – A research team in China has developed a breakthrough semiconductor chip that mimics the human brain's visual processing system, enabling robots and autonomous vehicles to detect motion four times faster than the human eye.

The study, led by Professor Gao from Beihang University and Professor Li Tao from the Beijing Institute of Technology, was published on February 10 in the prestigious international journal Nature Communications.

Mimicking the Human Brain
The core of this innovation lies in "neuromorphic computing"—designing hardware that functions like the biological neural networks of a brain.

In traditional computer vision, systems must analyze every single pixel of a video frame to detect movement. This process is computationally heavy; processing a standard high-definition image can take over 0.6 seconds—a delay that can be fatal for high-speed drones or self-driving cars.

The human brain, however, is much more efficient. When our retinas receive light, the lateral geniculate nucleus (LGN) acts as a relay station that filters out static backgrounds and highlights only the areas where movement occurs. The visual cortex then focuses its processing power solely on those moving parts.

A Leap in Reaction Speed
The research team replicated this process by arranging transistors in a grid-like lattice that mimics neural connections. These transistors respond to changes in brightness in just 0.1 milliseconds. By pre-filtering the image and isolating only the "moving zones," the chip allows software to ignore static data, drastically reducing the overall processing load.

The real-world implications of this speed increase are significant:

Autonomous Driving: In tests, the chip reduced image processing time by 0.2 seconds. For a vehicle traveling at 80 km/h, this translates to a 4.4-meter reduction in braking distance.
Drones: Reaction times were cut to less than one-third of existing speeds, allowing for flawless real-time tracking of erratic objects.
Robotics: The chip showed superior performance in high-speed scenarios, such as robotic arm manipulation and tracking ping-pong balls.

Durability and Efficiency
Beyond speed, the chip demonstrated impressive technical resilience. The researchers reported that the transistors could maintain information for over 10,000 seconds even without power and remained functional after more than 8,000 cycles of repeated use.

"This research brings video analysis speeds beyond human capabilities by applying the brain's visual principles to silicon," the research team stated. "We expect this to be a game-changer for collision avoidance in self-driving cars, real-time drone surveillance, and human-robot interaction where instant response to gestures is required."

[Copyright (c) Global Economic Times. All Rights Reserved.]