Machine learning based canine posture estimation using inertial data

Abstract

The aim of this study was to design a new canine posture estimation system specifically for working dogs. The system was composed of Inertial Measurement Units (IMUs) that are commercially available, and a supervised learning algorithm which was developed for different behaviours. Three IMUs, each containing a 3-axis accelerometer, gyroscope, and magnetometer, were attached to the dogs’ chest, back, and neck. To build and test the model, data were collected during a video-recorded behaviour test where the trainee assistance dogs performed static postures (standing, sitting, lying down) and dynamic activities (walking, body shake). Advanced feature extraction techniques were employed for the first time in this field, including statistical, temporal, and spectral methods. The most important features for posture prediction were chosen using Select K Best with ANOVA F-value. The individual contributions of each IMU, sensor, and feature type were analysed using Select K Best scores and Random Forest feature importance. Results showed that the back and chest IMUs were more important than the neck IMU, and the accelerometers were more important than the gyroscopes. The addition of IMUs to the chest and back of dog harnesses is recommended to improve performance. Additionally, statistical and temporal feature domains were more important than spectral feature domains. Three novel cascade arrangements of Random Forest and Isolation Forest were fitted to the dataset. The best classifier achieved an f1-macro of 0.83 and an f1-weighted of 0.90 for the prediction of the five postures, demonstrating a better performance than previous studies. These results were attributed to the data collection methodology (number of subjects and observations, multiple IMUs, use of common working dog breeds) and novel machine learning techniques (advanced feature extraction, feature selection and modelling arrangements) employed. The dataset and code used are publicly available on Mendeley Data and GitHub, respectively.

Unsupervised Domain Adaptation for Mitigating Sensor Variability and Interspecies Heterogeneity in Animal Activity Recognition

“…Neck sensors are widely used in AAR due to their ease of attachment to animals [4][5][6][7][8][9][10][13][14][15][16][17]. However, neck sensors have shown a lower classification performance due to orientation shifts caused by the tightness of sensors [13,14,48]. Our findings in Figures 3 and 4 are consistent with previous studies, which also reported lower classification results when predicting animal activities with the neck sensor data.…”

“…Moreover, these sensors can be attached to various body parts depending on the purpose of the measurement, environmental conditions, and animal characteristics. Neck sensors are widely used in AAR due to their ease of attachment to animals [4][5][6][7][8][9][10][13][14][15][16][17]. However, neck sensors have shown a lower classification performance due to orientation shifts caused by the tightness of sensors [13,14,48].…”

“…One of the challenges in Animal Activity Recognition (AAR) is the variability introduced by different sensor positions. The back sensor is optimized for capturing larger movements, whereas the neck sensor is designed to detect intricate movements [13,14]. This variability is further exacerbated by differences in experimental settings and labeling criteria, which can diverge significantly among researchers [13][14][15].…”

“…The back sensor is optimized for capturing larger movements, whereas the neck sensor is designed to detect intricate movements [13,14]. This variability is further exacerbated by differences in experimental settings and labeling criteria, which can diverge significantly among researchers [13][14][15]. In recent years, researchers have made efforts to tackle individual or sensor variability by leveraging stateof-the-art machine learning techniques [16,17].…”

“…However, this approach still requires labeled data for each class to fine-tune the global model and customize individual models. Acquiring sufficient data for deep learning models is time-consuming, particularly given the unpredictable activities exhibited by animals, even those that are trained, such as dogs [13][14][15]. Consequently, the data collection and labeling process demands substantial domain knowledge, time, and resources, leading to limitations in the availability of large labeled datasets for supervised learning approaches [22].…”