Pose-gait analysis for cetacean biologging tag data.

Area of Science:

• Marine biology
• Biomechanics
• Data science

Background:

• Biologging tags are crucial for studying cetacean behavior and locomotion.
• Accurate gait parameterization from tag data is challenging due to variable tag orientation.
• Current methods rely on custom, species-specific scripts, impacting reliability and repeatability.

Purpose of the Study:

• To develop an automated data processing pipeline for estimating tag orientation relative to the animal's body.
• To enable robust and repeatable gait analysis from biologging tag data.
• To enhance the investigation of marine mammal movement and behavior.

Main Methods:

• Developed a data-driven pipeline to identify relative motion between tag and animal.
• Estimated tag orientation (roll, pitch, yaw) using gait patterns.
• Calculated pose-invariant gait parameters.
• Validated the method with bottlenose dolphin, humpback whale, and beluga whale tag data.

Main Results:

• Achieved an average relative orientation error (tag wrt body) within 11 degrees for roll, pitch, and yaw.
• Demonstrated high precision (0.87) and recall (0.89) in detecting relative motion in dolphin data.
• Successfully applied the gait analysis to enhance movement and behavior studies in whales.

Conclusions:

• The automated pipeline provides accurate and repeatable estimation of tag orientation.
• This method improves the analysis of gait parameters, offering insights into cetacean locomotion.
• The publicly available software and data facilitate broader application in marine mammal research.