Development of Embedded AI Applications & Toolchain

Abstract

The relentless advancement of Artificial Intelligence (AI) is revolutionizing numerous industries, notably the embedded systems domain. This report sheds light on the development and deployment of cutting-edge, embedded AI applications and an efficient tool chain, focusing on gesture recognition using radar sensor, developing machine learning models, and r&d on neural architecture search algorithm.

The first major achievement was the development and deployment of a radar gesture classification model on the Infineon XENSIV™ kit. This innovative model could identify four distinct gestures in real-time, serving as a testament to the power of embedded AI in enhancing interaction between humans and machines. With the ability to recognize and interpret specific gestures, the model provided a foundation for creating more intuitive and interactive user interfaces for a wide range of applications. In addition to radar gesture recognition, the internship involved the development of various machine learning models. These models were designed to harness the power of AI for solving complex tasks, demonstrating the capacity of machine learning in improving the efficiency of embedded systems. This work further highlighted the potential of machine learning models to advance data-driven solutions in the field of embedded AI. Lastly, the report delves into the research, exploration, and development of "Neural Architecture Search". This groundbreaking work involved the design of neural networks that are aware of the hardware they run on, potentially revolutionizing the way machine learning models are developed and deployed in real-world settings. By taking into consideration the hardware constraints during the design of neural networks, this research could pave the way for more efficient and practical AI solutions.

In summary, the work chronicled in this report underscores the transformative potential of AI in the realm of embedded systems. It offers valuable insights and contributions in radar gesture recognition, machine learning models, and hardware-aware neural architecture search, showcasing a promising future for the integration of AI into everyday devices.