TLBC is an R package for classifying human behaviors from accelerometer and/or GPS data using a two-level behavior classifier (TLBC). The package has been developed for Actigraph accelerometers placed on either the hip or wrist.
Models that have been pre-trained on three UCSD datasets are available to download.
Two research assistants wore an Actigraph GT3X+ on the hip and a GPS. The research assistants performed a specified sequence of activities around San Diego county. We have used their data to train models to recognize the following behaviors: riding in a car, standing, walking, riding in a bus, sitting, some movement, bicycling. There is a trained model for the hip GT3X+ accelerometer only, the GPS only, and both devices combined.
Forty commuter cyclists wore an Actigraph GT3x+ on the hip and a GPS and performed their usual daily activities for 2-3 days. Their behaviors were annotated from images collected from a wearable camera. We have used their data to train models to recognize the following behaviors: sitting, standing still, standing moving, walking/running, bicycling and riding in a vehicle. There is a trained model for the hip GT3X+ accelerometer only, the GPS only, and both devices combined.
Thirty-six overweight women wore an Actigraph GT3X+ on the hip and wrist and a GPS and performed their usual daily activities for 5-7 days. Their behaviors were annotated from images collected from a wearable camera. We have used their data to train models to recognize the following behaviors: sitting, standing still, standing moving, walking/running, bicycling and riding in a vehicle. There is a trained model for the hip GT3X+ accelerometer only, the wrist GT3X+ only, the GPS only, the hip GT3X+ and GPS combined, and the wrist GT3X+ and GPS combined.
© Katherine Ellis. | Design by TEMPLATED.