Identification of children's activity type with accelerometer-based neural networks.

PURPOSE: The study's purpose was to identify children's physical activity type using artificial neural network (ANN) models based on uniaxial or triaxial accelerometer data from the hip or the ankle.

METHODS: Fifty-eight children (31 boys and 27 girls, age range = 9-12 yr) performed the following activities in a field setting: sitting, standing, walking, running, rope skipping, playing soccer, and cycling. All children wore uniaxial and triaxial ActiGraph accelerometers on both the hip and the ankle. Four ANN models were developed using the following accelerometer signal characteristics: 10th, 25th, 75th, and 90th percentiles; absolute deviation; coefficient of variability; and lag-one autocorrelation. The accuracy of the models was evaluated by leave-one-subject-out cross-validation.

RESULTS: The models based on hip accelerometer data correctly classified the activities 72% and 77% of the time using uniaxial and triaxial accelerometer data, respectively, whereas the models based on ankle accelerometer data achieved a percentage of 57% and 68%. The hip models were better able to correctly classify the activities walking, rope skipping, and running, whereas the ankle models performed better when classifying sitting. The models based on triaxial accelerometer data produced a better classification of the activities standing, running, rope skipping, playing soccer, and cycling than its uniaxial counterparts.

CONCLUSIONS: Applying ANN models to processing accelerometer data from children is promising for classifying common physical activities. The highest percentage of correctly classified activities was achieved when using triaxial accelerometer data from the hip.