Knowledge engineering for unsupervised canine posture detection from IMU data

Training animals is a process that requires a significant investment of time and energy on the part of the trainer. One of the most basic training tasks is to train dogs to perform postures on cue. While it might be easy for a human trainer to see when an animal has performed the desired posture, it is much more difficult for a computer to determine this. Most work in this area uses accelerometer and/or gyroscopic data to produce data from an animal's current state, but this has limitations. Take for example a normal standing posture. From an accelerometer's perspective, it closely resembles the "laying down" posture, but the posture can look very different if the animal is standing still, versus walking, versus running, and might look completely different from a "standing on incline" posture. A human trainer can instantly tell the difference between these postures and behaviors, but the process is much more difficult for a computer. This paper demonstrates several algorithms for recognizing canine postures, as well as a system for building a computational model of a canine's potential postures, based solely on skeletal measurements. Existing techniques use labeled data, which can be difficult to acquire. We contribute a new technique for unsupervised posture detection, and compare the supervised technique to our new, unsupervised technique. Results indicate that the supervised technique performs with a mean 82.06% accuracy, while our unsupervised approach achieves a mean 74.25% accuracy, indicating that in some cases, our new unsupervised technique is capable of achieving comparable performance.