Registration Is Open! Early Bird Pricing Expires June 30th
ActiGraph Digital Data Summit 2021November 4 - 5 | Learn more
Identification of Children's Activity Type with Accelerometer-Based Neural Networks
- Added on June 16, 2011
Introduction Children’s physical activity has traditionally been measured with self-reports. Self-reports are easily administered, low-cost measurements. However, they do not capture the sporadic short-burst nature of children’s physical activity very well. Accelerometers have therefore, in recent times, become the method of choice in physical activity research. These devices provide objective information about the frequency, intensity, and duration of physical activity. However, they do not provide information about the type of physical activity children engage in. Pattern-recognition-based approaches have shown to be successful in classifying a number of controlled physical activities among adults and elderly. To our knowledge, studies applying these approaches to accelerometer data from children are currently lacking. The physical activity pattern of children is very different from that of adults. Children’s physical activity pattern is characterized by frequent spasmodic bursts of short duration. They participate in intermittent and unstructured activities and the type of activities children engage in changes as they develop, going from informal active play during early childhood to activities that begin to mirror those of adults during adolescence. Because of these differences, it remains to be seen whether activity classification based on accelerometer data is just as successful in children as in adults and elderly. The purpose of this study was to identify types of physical activity among school-aged children using artificial neural network models based on three-axial accelerometer data from the hip or the ankle. Secondarily, it was examined whether the accuracy of the developed ANN models improved by including more information about the intensity of activities by mean of heart rate data or by adding information about the velocity of activities from global positioning systems (GPS).
Methods Fifty-eight children (31 boys; 27 girls; age range: 9-12 years) performed the following activities in a field setting: sitting, standing, walking, running, rope skipping, playing soccer, and cycling. All children wore a three-axial ActiGraph accelerometer on both the hip and the ankle, a Qstarz Travel Recorder GPS device, and a Polar heart rate device. 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, and the mean of the heart frequency and the mean speed computed for 10 seconds intervals. The accuracy of the models was evaluated by leave-one-subject-out cross-validation.
Results The hip model based on accelerometer data correctly classified the activities 77% of the time while the ankle model achieved a percentage of 57. The hip model was better able to correctly classify the activities walking, rope skipping, and running, whereas the ankle model performed better when classifying sitting. Preliminary results show that the accuracy of both models improves when information about heart rate or speed are included. The highest improvement is achieved for the activities cycling and rope skipping.
Discussion and Conclusion Applying ANN models to process accelerometer data from children is promising for classifying common physical activities. The performance of models based on
accelerometer data from the hip is better than the performance of models based on accelerometer data from the ankle. Some activities can be better classified when including information about heart rate and speed in the models.
Reference  Vries SI de, Galindo Garre F, Engbers LH, Buuren S van. Evaluation of Neural Networks to Identify Types of Activity Using Accelerometers. Med Sci Sports Exerc, in press.
ICAMPAM- Glasgow 2011