The book presents cutting edge applications of artificial intelligence and deep learning to short-range radar. Artificial Intelligence (AI) is the hottest topic in all industry sectors and has led to disruptions across all fields, computer vision, natural language processing, speech processing, medical imaging, etc. However, the application of AI to radars is relatively new and unexplored. We in this book present the cutting edge applications that we worked and are working on at Infineon. The book covers applications ranging from industrial, consumer space to emerging automotive applications. The book presents several human-machine interface (HMI) applications such as gesture recognition and sensing, human activity classification, air-writing, material classification, vital sensing, people sensing, people counting, people localization and in-cabin automotive occupancy and smart trunk opening. Read on for a detailed chapter-by-chapter breakdown of the book:
Chapter 1 introduces radar signal processing, the various types of short-range radars, basics of waveform designs, target models for high resolution radars, 3D radar data-cube processing involving detection, parameter estimation and tracking algorithms. The chapter concludes with various advanced industrial, consumer and automotive applications of short-range radars.
Chapter 2 provides an introduction to deep learning, outlining the history of neural networks and the optimization algorithms to train them. The chapter introduces the modern deep convolutional neural network (DCNN), popular DCNN architectures for computer vision and their features. It presents other deep learning architectures such as long-short term memory (LSTM), auto-encoders, variational auto-encoders (VAE), and generative adversarial networks (GAN).
Chapter 3 introduces the application of gesture recognition. Particularly we divide them into macro-gesture sensing and recognition, which involve major arm movements and are performed from a distance to the sensor. And micro-gesture sensing and recognition, which involves minute subtle finger movements and are performed close to the sensor. While the macro-gesture finds application in projectors or TVs, micro-gesture finds applications in smartphones, smart-watches and automotive dash-boards.
Chapter 4 introduces the application of human activity recognition. The chapter presents various DCNN architectures for enabling activity recognition and comparing their performance. The chapter further introduces the concept of continuous activity recognition using a combination of tracking and deep learning classifier, such as LSTM. The chapter finally presents the topic of elderly fall motion recognition, the various challenges and how it can be addressed through deep learning.
Chapter 5 presents the application of air-writing using a network of short-range radars. Air-writing or air-drawing refers to linguistic characters drawn in imaginary board and the system should recognize these characters. Air-writing finds application in human-machine interface for augmented-reality virtual-reality (AR-VR) and alternate interface mechanisms for desktops.
Chapter 6 presents the application of material classification. The objective of short-range radar is to classify among everyday objects and materials, and finds applications in vacuum cleaners and robots. The chapter concludes with the challenges to be over-come for integration into deployable system solution.
Chapter 7 presents the application of remote vital sensing using short-range radars. The chapter presents various challenges with pure deep learning solution and introduces tracking based solution for a stable solution, and further introduces a hybrid tracking with deep learning solution for a much reliable and scalable solution.
Chapter 8 introduces the application of short-range radars in conjunction to deep learning for human presence, counting and localization applications. The chapter presents how deep learning has and is replacing classical signal processing pipeline to improve the system performance as well as addressing challenges which weren’t feasible with classical signal processing pipeline.
Chapter 9 introduces the application of in-cabin sensing. Apart from automatic cruise control (ACC), blind spot detection (BSD), and parking assistance, radars are enabling plethora of in-cabin sensing applications such as smart trunk opening (STO) applications, child-left-behind applications and smart air-bags applications. We conclude the chapter with federated learning framework, wherein a mechanism to automatically update a deployed deep learning model in production is outlined.
Click here for more information or to order.
