Dog behaviour and stress monitoring in a pet shelter kennel (together with DierENopvang Hengelo)
Problem Statement:
Pet shelters provide a temporary home for dogs, often housing them in kennels designed to ensure safety, comfort, and socialization. However, the first few nights in a shelter can be stressful for dogs [1], as stress—manifested through behaviours such as excessive barking, pacing, or changes in activity levels—arises due to unfamiliar environments and stimuli. Monitoring stress is essential for promoting the well-being and adjustment of dogs, with day and night behavioural observations offering valuable insights into their acclimation. Traditional observation methods, however, are labour-intensive, prone to human error, and risk altering the dogs’ natural behaviours due to human presence. To address these limitations, unobtrusive sensing technologies, such as Wi-Fi [2] and infrared systems, provide a promising alternative for long-term, consistent stress monitoring. These methods operate passively, minimize interference with natural behaviour, and generate comprehensive data for analysis, enabling an improved understanding of canine stress and well-being in shelter environments.
[1] van der Laan, J.E., Vinke, C.M. & Arndt, S.S. Nocturnal activity as a useful indicator of adaptability of dogs in an animal shelter and after subsequent adoption. Sci Rep 13, 19014 (2023). https://doi.org/10.1038/s41598-023-46438-9
[2] Klein Brinke, J. (2024). Interwoven Waves: Enhancing the Scalability and Robustness of Wi-Fi Channel State Information for Human Activity Recognition. [PhD Thesis - Research UT, graduation UT, University of Twente]. University of Twente. https://doi.org/10.3990/1.9789036561419
Task:
Your tasks will cover the entire (Master-level) or part of (Bachelor-level) the implementation stack. The goal of this assignment is to cover anything from hardware design/implementation to data collection and analysis to data visualization and feedback systems. Your tasks likely include some form of data collection and analysis from different unobtrusive sensors (radio waves, thermal, audio, etc) and analysis through deep learning, with a focus on embedded AI (as ideally, the reasoning happens locally on cost and energy-efficient ESP32s or Raspberry Pi). You can pick your technology yourself! :)
Example RQS:
- How accurately can activity levels and stress indicators in shelter dogs be quantified using unobtrusive sensors (infrared, mmWave, audio, and thermal video) during the acclimation period?
- What is the impact of on-device embedded AI models on the accuracy and latency of stress detection in shelter dogs using ESP32 or Raspberry Pi platforms?
- How can audio, thermal video, and radio wave sensors be optimized to filter out human-related data while maintaining accuracy in monitoring stress-related behaviours in shelter dogs, in compliance with GDPR?
What am I looking for:
I am looking for motivated students who go beyond the 'sixes' attitude, ideally motivated to publish their results in the future. Additionally, you should be affectionate and not scared of dogs for this assignment, as there is a practical aspect to this assignment: Dierenopvang Hengelo has kindly opened its doors for unobtrusive data collection (meaning your devices need to be dog-secure).
Due to the flexibility of the assignment, I wish to have a coffee with you to ensure it’s a good match before I agree to supervise! ;)
Work:
40% implementation (hardware/software), 40% Data analysis, 20% Writing
Contact:
Jeroen Klein Brinke (j.kleinbrinke@utwente.nl)
Allow multiple students: YES
Photo: https://isorepublic.com/photo/cute-dog/