Activity Recognition Using a Single Accelerometer Placed at the Wrist or Ankle

摘要: AB Purpose: Large physical activity surveillance projects such as the UK Biobank and NHANES are using wrist-worn accelerometer-based monitors that collect raw data. The goal is to increase wear time by asking subjects on wrist instead of hip, then use information in signal improve type intensity estimation. purposes this work was obtain an algorithm process ankle data classify behavior into four broad classes: ambulation, cycling, sedentary, other activities. Methods: Participants (N = 33) wearing accelerometers performed 26 daily accelerometer were collected, cleaned, preprocessed extract features characterize 2-, 4-, 12.8-s windows. Feature vectors encoding about frequency motion extracted from analysis used with a support vector machine classifier identify subject's activity. Results compared categories classified human observer. Algorithms validated leave-one-subject-out strategy. computational complexity each processing step also evaluated. Results: With windows, proposed strategy showed high classification accuracies for (95.0%) decreased 84.7% Shorter (4 s) windows only minimally performances 84.2%. Conclusions: A 13 shows good classes given activities original set. computationally efficient could be implemented real mobile devices 4-s latency.