Project description
AI-based portable radar system to track human activity
AI-powered algorithms enable radar systems to perform human activity recognition (HAR). However, previous bulky radar systems consumed substantial power, rendering them unsuitable for indoor security, healthcare and robotics applications. To address this challenge, the MSCA-funded AIRHAR project aims to develop a portable radar system for HAR. This system employs a hardware–software co-design approach to significantly reduce energy consumption for signal processing compared to traditional radar-based HAR systems. The project is focused on enhancing the efficiency of radar data analysis, with a specific emphasis on feature extraction to achieve a 50 % reduction in energy consumption. Additionally, it will incorporate neuromorphic principles to reduce the cost of deep neural networks and develop AI accelerator hardware to achieve a remarkable 10x reduction in energy consumption. These results have the potential to revolutionise radar data processing and significantly contribute to the reduction of greenhouse gas emissions.
Objective
Radar systems have been used in ambient sensing to track various subjects using electromagnetic waves. Thanks to the increasing capability of artificial intelligence algorithms in solving classification tasks, human activity detection (HAR) using radar systems have become possible. However, most previous solutions use bulky fixed radar systems with tens to hundreds of watts of power consumption, requiring rigid wall plug connection, making them environmentally unfriendly and difficult to use in applications like indoor security, healthcare, and mobile robots.
In this project, we aim to develop a portable radar system for HAR by following a hardware-software co-design approach to significantly reduce the signal processing energy consumption compared to conventional radar-based HAR systems. On the software side, we will explore novel time-domain feature extraction methods to reduce the energy consumption of radar data analysis by at least 2 times. We will also apply brain-inspired neuromorphic principles to reduce 50 times the energy cost of state-of-the-art deep neural network architectures to solve radar-based HAR tasks. On the hardware side, we will develop artificial intelligence (AI) accelerator hardware based on field-programmable gated arrays (FPGAs) and application-specific integrated circuits (ASICs) to decrease the hardware energy consumption in radar data processing by at least 10 times.
Overall, we expect the project results to revolutionize the paradigm of radar data processing to achieve over 20 times the whole system energy reduction, which will significantly contribute to the target of greenhouse gas emission reduction defined in the European Green Deal. The developed energy-efficient AI-powered portable radar system for HAR is promising to initiate a commercialized product for indoor security, healthcare, and mobile robot applications, and it will be competitive in the rapidly growing Internet-of-Things market worth USD 2400 billion by 2029.
Fields of science (EuroSciVoc)
CORDIS classifies projects with EuroSciVoc, a multilingual taxonomy of fields of science, through a semi-automatic process based on NLP techniques.
CORDIS classifies projects with EuroSciVoc, a multilingual taxonomy of fields of science, through a semi-automatic process based on NLP techniques.
- natural sciencescomputer and information sciencessoftware
- engineering and technologyelectrical engineering, electronic engineering, information engineeringinformation engineeringtelecommunicationsradio technologyradar
You need to log in or register to use this function
Keywords
Programme(s)
- HORIZON.1.2 - Marie Skłodowska-Curie Actions (MSCA) Main Programme
Funding Scheme
HORIZON-TMA-MSCA-PF-EF - HORIZON TMA MSCA Postdoctoral Fellowships - European FellowshipsCoordinator
2628 CN Delft
Netherlands