REAL-TIME DETECTION OF THE ACTIVITY OF A DOG

This paper introduces our preliminary work with assistance dogs. Even when dogs are very well trained some problems may occur in practice, typical examples are dog escaping or running after a cat. Our long term objective is to take advantage of the technology to increase the safety of the dog and its owner. Our first work focuses on the activity classification of the dog. This paper presents preliminary results for recognizing four types of activity: walk, run, lay and sit down. Experiments and results on real data collected with low-cost gyroscopes and accelerometers are presented and discussed.