SpokeSense: Developing a Real-Time Sensing Platform for Wheelchair Sports
Patrick Carrington, Gierad Laput, Jeffrey P. Bigham · 2020 · SIGACCESS Access. Comput. · doi:10.1145/3386308.3386310
Summary
This paper presents SpokeSense, a wheel-mounted sensing platform designed to track and analyze wheelchair basketball performance in real time. While substantial work has gone into sensing and analytics for mainstream sports, wheelchair sports have received relatively little attention despite their growing popularity. Wheelchair sports present unique sensing challenges: athletes do not take steps, wrist-worn devices are impractical during play, and the performance metrics that matter differ from able-bodied sports. SpokeSense uses two custom sensors called Spokes, mounted in hubcap-style housings over each wheel hub. Each Spoke contains a Particle Photon microcontroller, a 9-DOF inertial measurement unit (IMU) for motion tracking, a MEMS microphone for detecting in-game audio events, and a LiPo battery — all on a compact custom PCB measuring 1.5 x 1.73 inches. The sensors stream data over WiFi to a host computer that processes and visualizes the information. The system provides two levels of feedback: core metrics (speed, distance, acceleration) and mid-level insights including speed zones (six zones from non-motion to very high, relative to the athlete's maximum speed), orientation change events (detecting intentional tilts versus dangerous tipping), and contextual audio clues (recognizing dribbling, referee whistles, and game buzzers).
Key findings
Accuracy testing with 21 participants — including 7 non-players, 1 wheelchair basketball player, and 13 National Wheelchair Basketball Tournament players — showed the system achieves less than 5% error for speed and distance measurement in lab conditions (0.57% absolute error), with field test error at 13.41% due to surface variations and real-world conditions. The audio event detection system, built using a CNN-based classifier trained on game recordings from a Division I championship game, achieved 98.9% clip-level accuracy for detecting whistles, dribbling, and buzzers. Even when tested with adversarial foreign sounds (people chattering, baby crying, phone ringing), accuracy remained at 98.1%, demonstrating robustness. The paper identifies five open challenges for wheelchair sports sensing: supporting data sharing between athletes with similar abilities, optimizing the platform across different wheelchair sports, ensuring athlete values and identities are not sacrificed in pursuit of generalized solutions, integrating with existing team workflows, and navigating clearance for in-game use.
Relevance
SpokeSense represents an important contribution to accessible sports technology, an area that is often overlooked in both assistive technology research and mainstream sports analytics. The work challenges the assumption that fitness and performance tracking designed for able-bodied athletes can simply be adapted for wheelchair users — instead demonstrating that wheelchair sports require purpose-built sensing solutions. The open challenges raised are particularly thoughtful: the authors caution against normative approaches to physical activity monitoring that could impose idealized performance standards on athletes with disabilities, advocating instead for personalization that supports individuality. For accessibility practitioners, this research highlights how assistive technology extends well beyond productivity and communication into recreation, sport, and quality of life — domains where people with disabilities deserve equal access to the data-driven tools that have transformed mainstream athletics.
Tags: wheelchair sports · adaptive sports · wheelchair basketball · wearable technology · sensors · physical activity · assistive technology · sports analytics