Temporal segmentation of exercise movements from wearable sensors

多普尔公司是一家总部位于安大略省滑铁卢的加拿大初创公司，专注于可穿戴健身的开发 运动训练用具。多佩尔正在开发一种可穿戴的健身设备，可以跟踪CrossFit的重复动作 运动员。为了促进班级内的合规、参与和竞争，结果是 实时显示在记分板上，并可供用户在线访问。多佩尔已经开发出一种 基于带状氯丁橡胶内置IMU的运动过程在线跟踪装置 连接在胸部水平的频带(就像心率监测器)，它通过蓝牙将数据传输到中央 设备。为了能够在线统计和显示每个参与者的练习重复次数， 系统必须能够准确地识别每次重复的起点和终点 IMU测量的时间序列数据，这是一个称为时间分割的问题。建议的目的是 研究的目的是开发和验证一种准确且健壮的分割算法，该算法能够 准确生成CrossFit练习所需的重复数据。

Doppel Inc. is a Canadian startup based in Waterloo, Ontario, focused on the development of wearable fitness devices for athletic training. Doppel is creating a wearable fitness device that tracks repetitions for CrossFit athletes. In order to promote compliance, engagement and competition within the class, the results are displayed on a scoreboard in real time, and made available online for users to access. Doppel has developed a device for the on-line tracking of movement during exercise, based on an IMU built into a belt-like neoprene band, attached at the chest level (like a heart rate monitor), which transmits data via bluetooth to a central device. In order to enable the online counting and display of exercise repetitions for each participant, the system must be able to accurately identify the starting and ending point of each repetition from the continuous time series data measured by the IMU, a problem known as temporal segmentation. The aim of the proposed research is to develop and validate an accurate and robust segmentation algorithm which is capable of accurately generating the required repetition data for the CrossFit exercises.