A&S Graduate Studies
[PAST EVENT] [via Zoom] Amanda Watson, Computer Science - Ph.D. Defense
Abstract:
Wearable technology research has led to advancements in healthcare and athletic performance. Devices range from one size fits all fitness trackers to custom fitted devices with tailored algorithms. Because these devices are comfortable, discrete, and pervasive in everyday life, custom solutions can be created to fit an individual’s specific needs. In this dissertation, we design wearable sensors, develop features and algorithms, and create intelligent feedback systems that promote the advancement of healthcare and athletic performance.
First, we present Magneto: a body mounted electromagnet-based sensing system for joint motion analysis. Joint motion analysis facilitates research into injury prevention, rehabilitation, and activity monitoring. Sensors used in such analysis must be unobtrusive, accurate, and capable of monitoring fast-paced dynamic motions. Our system is wireless, has a high sampling rate, and is unaffected by outside magnetic noise. Magnetic noise commonly influences magnetic field readings via magnetic interference from the Earth’s magnetic field, the environment, and nearby ferrous objects. Magneto uses the combination of an electromagnet and magnetometer to remove environmental interference from a magnetic field reading. We evaluated this sensing method to show its performance when removing the interference in three movement dimensions, in six environments, and with six different sampling rates. Then, we localized the electromagnet with respect to the magnetic field reader, allowing us to apply Magneto in two pilot studies: measuring elbow angles and calculating shoulder positions. We calculated elbow angles to the nearest 15? with 93.8% accuracy, shoulder position in two-degrees of freedom with 96.9% accuracy, and shoulder positions in three-degrees of freedom with 75.8% accuracy.
Second, we present TracKnee: a sensing knee sleeve designed and fabricated to unobtrusively measure knee angles using conductive fabric sensors. We propose three models that can be used in succession to calculate knee angles from voltage. These models take an input of voltage, calculate the resistance of our conductive fabric sensor, then calculate the change in length across the front of the knee and finally to the angle of the knee. We evaluated our models and our device by conducting a user study with six participants where we collected 240 ground truth angles and sensor data from our TracKnee device. Our results show that our model is 94.86% accurate to the nearest 15th degree angle and that our average error per angle is error per angle is 3.69 degrees.
Third, we present ServesUp: a sensing shirt designed to monitor shoulder and elbow motion during the volleyball serve. In this project, we will designed and fabricated a sensing shirt that is comfortable, unobtrusive, and washable that an athlete can wear during and without impeding volleyball play. To make the shirt comfortable, we used soft and flexible conductive fabric sensors to monitor the motion of the shoulder and the elbow. We conducted a user study with ten volleyball players for a total of 1000 volleyball serves. We classified serving motion using a KNN with a classification accuracy of 89.2%. We will use this data provide actionable insights back to the player to help improve their serving skill.
Fourth, we present BreathEZ, the first smartwatch application that provides both choking first aid instruction and real-time tactile and visual feedback on the quality of the abdominal thrust compressions. We evaluated our application through two user studies involving 20 subjects and 200 abdominal thrust events. The results of our study show that BreathEZ achieves a classification accuracy of 90.9% for abdominal thrusts. All participants that used BreathEZ in our study were able to improve their performance of abdominal thrusts. Of these participants, 60% were able to perform within the recommended range with the use of BreathEZ. Comparatively no participants trained with a video only reached that range.
Finally, we present BBAid: the first smartwatch based system that provides real-time feedback on the back blow portion of choking first aid while instructing the user on first aid procedure. We evaluated our application through two user studies involving 26 subjects and 260 back blow events. The results of our study show that BBAid achieves a classification accuracy of 93.75% for back blows. With the use of BBAid, participants in our study were able to perform back blows within the recommended range 75% of the time. Comparatively the participants trained with a video only reached that range 12% of the time. All participants in the study, after receiving training were much more willing to perform choking first aid.
Bio:
Amanda Watson is a Ph.D. candidate in the Computer Science Department at William & Mary where she researches under Dr. Gang Zhou in the LENS lab. She received her M.S. in Computer Science from William & Mary and a B.A. in Computer Science and Mathematics from Drury University where she was a 4-year letterman on the varsity volleyball team. Her research interests include Wearable Technology for Healthcare and Athletic Performance.