Imec has unveiled a radar-signal processing chip that uses a spiking recurrent neural network (SNN). According to the Leuven-based innovation center, the new IC, which mimics biological neurons in pattern recognition, consumes 100 times less power than traditional implementations and offers a tenfold reduction in latency – allowing for near-real-time decision-making. The chip is slated for initial use as part of a low-power, highly intelligent anti-collision radar system for drones.
Artificial neural networks (ANNs) are a key ingredient of radar-based anti-collision systems commonly used in the automotive industry. However, Imec sees nearly limitless potential for its new development. In fact, this chip was initially designed to support electrocardiogram (ECG) and speech processing in power-constrained devices. However, with its generic architecture and new digital hardware design, its algorithms can easily be fine-tuned to process a variety of sensor data – including electrocardiogram, speech, sonar, radar and lidar streams. Contrary to analog SNN implementations, Imec’s event-driven digital design makes the chip behave exactly and repeatedly as predicted by the neural network simulation tools.
“SNNs operate very similarly to biological neural networks, in which neurons fire electrical pulses sparsely over time, and only when the sensory input changes. As such, energy consumption can significantly be reduced,” describes Ilja Ocket, program manager of neuromorphic sensing at Imec. “What’s more, the spiking neurons on our chip can be connected recurrently – turning the SNN into a dynamic system that learns and remembers temporal patterns. The technology we’re introducing today is a major leap forward in the development of truly self-learning systems.”