Contact-Free Simultaneous Sensing of Human Heart Rate and
Canine Breathing Rate for Animal Assisted Interactions[Bibtex]

Application of Computer vision
to Estiamte Canine Breathing Rate and Human Heart Rate from Video in AAI context

Timothy Holder 1
Mushfiqur Rahman1
Emily Summers1
David Roberts1
Chau-Wai Wong1
Alper Bozkurt1
1Noth Carolina State University

>Full Code [GitHub] [Colab]

ACI 2022 [Paper]




Abstract

Animal Assisted Interventions (AAIs) are interactions between people and animals, i.e., patting a dog, and those can benefit both. Analyzing the outcome in human or animal is challenging due to subjective nature of the outcome and difficulties in attaching sensors to animals. This paper proposes a new approach that uses cameras to remotely measure human heart rate and dog breathing rate during interactions, and shows promising results. This method could have applications in various fields beyond research, including veterinary, surgical, and clinical settings.

Video Tutorial on the Published Code

Presentation Video

9th International Conference on Animal-Computer Interaction, 5-8 December 2022, UK
Presented by Timothy Holder and Mushfiqur Rahman

Result
Discussion

In this study, we introduced a contact-free method using regular cameras to study interactions between humans and dogs. We remotely measured the heart rate of humans and the breathing rate of dogs during their interactions. We recorded videos with an iPhone Camera to capture the intensity of human face and hands that fluctuates periodically due to the change of oxygen concentrataion in blood. Delicate analysis of spectrogram in Fourier domain is necessary to separate subtle heart rate signal from other noise components. The camera estimated heart rate is compared with gold statndard heart rate such as that from wrist watch. For breathing rate estimation we extract the flow of motion in x and y direction along the region of a dog's belly. A neural netowrk is used to capture the motion and then the motion is analyzed in the frequency domain similar to the heart rate estimation.

For delicate frequency domain analysis, we use the Adaptive multi-trace carving (AMTC) algorithm, that helps us to separate a weak frequency signal among unwanted strong ones.

For example, using AMTC we are able to separate out the breathing frequency of the animal from the frequency of periodic patting by human during AAI.

Citation


Timothy Holder, Mushfiqur Rahman, Emily Summers, David Roberts, Chau-Wai Wong, Alper Bozkurt,
“Contact-free simultaneous sensing of human heart rate and canine breathing rate for animal assisted interactions,”
Newcastle upon Tyne, UK, 5–8 Dec. 2022. (Paper)
[Bibtex]


Acknowledgments

We are thankful to the authors of Global Motion Aggregation and Adaptive Multi-Trace Carving for Robust Frequency Tracking in Forensic Applications. Source code of this webpage is adapted from Peter Wang.