Edge Processing and AI
The trend is for the “thing” in the Industrial Internet of Things, or the functionality at the heart of any medical device, to get smarter. To meet this demand, we work with a range of IP partners to ensure we select and deliver the most appropriate processor core or DSP to meet each device’s specific requirements. There is also an increasing trend for IoT devices to exploit Artificial Intelligence (AI) with the use of Machine Learning (ML) algorithms. Convolutional Neural Networks (CNN) are among the most popular ML networks, which require a high level of matrix multiplications, for which DSP SIMD operations are a good fit. Dedicated Neural Network Accelerators (NNAs) are the most efficient, achieving 10x performance levels compared to a DSP, but less flexible. Recent advances in AI/ML have led to a range of NNAs being developed by our IP partners which can be integrated into edge devices for smart processing of data from audio, video, Radar, vibration or motion sensors. IP suppliers for NNA include Arm, Imagination Technologies, CEVA and Tensilica, our engineers have the skills and experience necessary to integrate these IPs into a SoC design alongside a conventional processor, DSP and/or GPU core.