ImeC Introduces First Spiking Neural Network Radar Chip, Cutting Power by 100× and Latency by 10×

ImeC has announced the launch of the world’s first spiking neural network (SNN) chip designed for radar signal processing. The breakthrough enables low‑power, high‑speed anti‑collision systems for drones that can detect and react to approaching objects within milliseconds.

By emulating the sparse firing patterns of biological neurons, the chip consumes 100 times less power than conventional solutions while reducing latency by a factor of ten, allowing almost instantaneous decision‑making. For instance, the chip can classify micro‑Doppler radar signatures using only 30 mW of power. Its versatile architecture can be adapted to process a wide range of sensor streams—ECG, speech, sonar, radar, and lidar—making it ideal for next‑generation intelligent systems.

Traditional artificial neural networks (ANNs) dominate many AI applications, including automotive radar‑based collision avoidance. However, ANNs often demand high power and require data to traverse the sensor–algorithm pipeline before a decision can be made, which is problematic for battery‑constrained devices.

“Today, we present the world’s first chip that processes radar signals using a recurrent spiking neural network,” says Ilja Ocket, program manager of neuromorphic sensing at ImeC. “SNNs operate in a way that closely mirrors biological neural networks, with neurons firing electrical pulses only when sensory input changes. This sparse activity dramatically reduces energy consumption. Moreover, the recurrent connections give the SNN the ability to learn and retain temporal patterns, marking a significant step toward truly self‑learning systems.”

ImeC’s chip was originally tailored for ECG and speech in ultra‑low‑power devices. Its generic, fully digital, event‑driven architecture allows straightforward reconfiguration for other modalities such as sonar, radar, and lidar. Unlike analog SNNs, the digital design guarantees that the chip’s behavior precisely matches the neural‑network simulation.

Smart low‑power anti‑collision system for drones (and cars)

A primary target for the new ImeC chip is a low‑latency, low‑power anti‑collision system for drones. The unmanned aerial vehicle industry—more so than automotive—operates with strict battery limits and requires rapid environmental awareness. By processing radar data on‑board, the chip can distinguish approaching objects faster and more accurately, enabling drones to respond to hazards almost instantly.

Ocket notes that one scenario under development involves autonomous drones navigating warehouses, maintaining safe distances from walls and shelves while performing complex tasks. The same technology could extend to robotics, automated guided vehicles, and even health monitoring.

To realize this chip, ImeC assembled a multidisciplinary team—from training‑algorithm developers and SNN architects inspired by neuroscience, to biomedical and radar signal‑processing experts, and ultra‑low‑power digital‑chip designers—ensuring a robust, end‑to‑end solution.

>> This article was originally published on our sister site, EE Times Europe.