CSR: Small: Wireless Systems as Gesture Sensors

University Of Washington, Seattle WA

Investigators

Abstract

Over the past two decades there has been a rapid proliferation of wireless signals, such as Wi-Fi and cellular, that support mobile devices. As a result, wireless signals have become ubiquitous in our environment -- at home, at work, and now even on airplanes. The goal of this project is to leverage these signals around us to enable novel sensing applications, such as gesture recognition. A gesture recognition system allows users to control devices using movements of their hands, arms, legs, etc. This research focuses on using wireless signals to achieve gesture-sensing capabilities that are currently impossible, particularly (1) gesture recognition in the home without the need for specific movement detectors (such as cameras or user-held devices), and (2) gesture recognition that can operate on battery-free devices. The project will address fundamental questions about the feasibility and limits of using existing wireless signals for extracting rich sensing information, and will explore deep connections to both communication and radar theory. This work has the potential to inspire the design of novel systems that can open up new research questions in multiple domains, including ubiquitous computing, mobile systems, computer networking, and human-computer interaction. The proposed inter-disciplinary research takes a novel approach of leveraging existing wireless systems to enable novel applications such as gesture sensing and recognition. The project will produce algorithms, circuits, sensors, designs, and system implementations to achieve this goal. The project will also be developing battery-free gesture sensors that can work using only harvested power from ambient wireless sources like TV and cellular transmissions. Further, these designs will be integrated with existing mobile devices such as smartphones and tablets to demonstrate always-on sensing on these power-constrained devices. The project will also develop classification algorithms that can run entirely on low-power micro-controllers without consuming significant amounts of power. Further, this research will lay the algorithmic foundations for extracting information from communication signals optimized for high-throughput (e.g., OFDM-based Wi-Fi and cellular transmissions) to enable rich sensing capabilities like whole-home gesture recognition and battery-free gesture recognition. Finally, the project demonstrate two main applications: reliable pointing gesture recognition (where the user points along different directions to activate devices) and through-the-pocket gesture recognition where the user can gesture at the phone in a pocket, e.g., to change volume or mute the phone.