CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Recognition of walking periods using various personal digital devices Authors:  Marcin Straczkiewicz - Harvard TH Chan School of Public Health (United States) [presenting]
Abstract: The development of body-worn devices, such as smartphones, smartwatches, and wearable accelerometers, has remarkably deepened our understanding of how physical activity impacts human health. However, these findings are possibly just a foretaste of what data of personal digital devices may reveal as many questions on their processing and analysis remain unanswered. One such question regards activity recognition. We will introduce a novel method for quantification of walking periods from various wearable devices. We utilize the temporal dynamics of body motions measured by accelerometer. We focus on salient features of walking, namely intensity, periodicity, duration, and speed. We investigate the reflection of these phenomena at several body locations typical to wearable devices, and we create a classification scheme that allows for flexible and interpretable estimation of walking. To assess the performance of our method, we validate it over more than 300 subjects from 14 publicly available physical activity datasets. We will demonstrate that our method achieves very high classification accuracy for the recognition of walking and does not overestimate walking during other common everyday activities, regardless of sensor placement and measurement parameters.