Vision and Sensing of Human Motion for Athletes and Artists

点击翻译按钮获取中文摘要

Biometrics typically refer to the application of pattern recognition methods to the measurement of humans for security purposes. We take a broader view, by applying a variety of signal processing and computer vision techniques to the measurement and analysis of human motion, for applications in athletics and clinical treatment of gait disorders. The prototypical application is an athlete running on a treadmill, but experiencing pain. We know that if the athlete runs differently, we can eliminate the pain, but gaits are not easily altered. Using athletics such as this as a case study, we seek to develop algorithms and systems that measure and analyze human motion to produce feedback to that person in real-time. A working hypothesis is that properly designed feedback will be an effective way to correct a gait. We have demonstrated the concept in previous work with a system designed to correct the technique of a speed skater who had lost his ability to do cross overs. We focus on audio feedback because it leaves the eyes free for the athlete's safety, and because we know that people can coordinate their movements with sound. The proposed research builds on this previous success by exploring the use of various sensors to measure locomotion (computer vision, motion capture, and sensors mounted on the athlete), measuring in a variety of settings (speed skating oval, treadmill, running track, sidewalk), developing signal processing algorithms for synchronization with the various sensors and motions, examining a variety of modes of locomotion (walking, running, skating), and developing and testing algorithms and systems that target a variety of desired outcomes (athletic performance, clinical treatment).