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

In Shift and In Variance: Assessing the Robustness of HAR Deep Learning Models Against Variability

Deep learning (DL)-based Human Activity Recognition (HAR) using wearable inertial measurement unit (IMU) sensors can revolutionize continuous health monitoring and early disease prediction. However, most DL HAR models are untested in their robustness to real-world variability, as they are trained on limited lab-controlled data. In this study, we isolated and analyzed the effects of the subject, device, position, and orientation variabilities on DL HAR models using the HARVAR and REALDISP datasets. The Maximum Mean Discrepancy (MMD) was used to quantify shifts in the data distribution caused by these variabilities, and the relationship between the distribution shifts and model performance was drawn. Our HARVAR results show that different types of variability significantly degraded the DL model performance, with an inverse relationship between the data distribution shifts and performance. The compounding effect of multiple variabilities studied using REALDISP further underscores the challenges of generalizing DL HAR models to real-world conditions. Analyzing these impacts highlights the need for more robust models that generalize effectively to real-world settings. The MMD proved valuable for explaining the performance drops, emphasizing its utility in evaluating distribution shifts in HAR data.

Rumination Detection in Sheep: A Systematic Review of Sensor-Based Approaches

The use of sensors to analyze behavior in sheep has gained increasing attention in scientific research. This systematic review aims to provide an overview of the sensors developed and used to detect rumination behavior in sheep in scientific research. Moreover, this overview provides details of the sensors that are currently commercially available and describes their suitability for sheep based on the information provided in the literature found. Furthermore, this overview lists the best sensor performances in terms of achieved accuracy, sensitivity, precision, and specificity in rumination detection, detailing, when applicable, the sensor position and epoch settings that were used to achieve the best results. Challenges and areas for future research and development are also identified. A search strategy was implemented in the databases PubMed, Web of Science, and Livivo, yielding a total of 935 articles. After reviewing the summaries of 57 articles remaining following filtration (exclusion) of repeated and unsuitable articles, 17 articles fully met the pre-established criteria (peer-reviewed; published between 2012 and 2023 in English or German; with a particular focus on sensors detecting rumination in sheep) and were included in this review. The guidelines outlined in the PRISMA 2020 methodology were followed. The results indicate that sensor-based systems have been utilized to monitor and analyze rumination behavior, among other behaviors. Notably, none of the sensors identified in this review were specifically designed for sheep. In order to meet the specific needs of sheep, a customized sensor solution is necessary. Additionally, further investigation of the optimal sensor position and epoch settings is necessary. Implications: The utilization of such sensors has significant implications for improving sheep welfare and enhancing our knowledge of their behavior in various contexts.