Abstract

Embedded systems have traditionally faced limitations in processing speed and adaptability, particularly in real-time applications. Advances in neural network acceleration offer a potential solution to these constraints. Current embedded systems often struggle to handle dynamic workloads efficiently, impacting performance in time- sensitive applications. There is a need for a novel approach to enhance processing capabilities without compromising real-time responsiveness. This study introduces a novel Adaptive Neural Acceleration Unit (ANAU) designed for 64-bit embedded systems. The ANAU leverages adaptive neural networks to dynamically adjust processing resources based on workload characteristics. The algorithm was implemented on a state-of-the-art embedded platform and evaluated across various real-time applications. The ANAU demonstrated a 35% increase in processing speed and a 40% reduction in power consumption compared to traditional methods. Real-time task latency improved by 25%, with system stability maintained under high- load conditions.