Automatically Quantifying Human Activity in Design Activities - With Microsoft Kinect Depth Sensors
MetadataVis full innførsel
Observing design activities can give insights that can be used to improve the activities and give feedback to participants. In the past the process of observing and analyzing design activities was done manually with video which is very resource demanding. With the launch of the Microsoft Kinect 3D depth sensor and later an improved version in 2014, automatically tracking human behavior became much easier. The thesis evaluates the suitability for the Kinect sensor to contribute to automating design observations and give real-time feedback to participants. A thorough analysis of the Kinect v1 and v2 sensors is conducted. A proof of concept for an observation setup and a sensing application is developed. The sensing application tracks participants from a ceiling mounted position and estimates real-time mechanical energy-use and creates heat maps of their movements. A prototype of a design experiment is developed to test the application and observation setup. The results show that, except for some challenges inherent to the depth sensing technology, the Kinect v2 sensor shows great potential for contributing to automating design observation setups. Finally, issues to be aware of when using Kinects specifically in design observation and Design Observatories are highlighted. Future work involves developing a permanent design observatory with Kinect sensors and exploring how the audio capabilities and gesture analysis software for the Kinect can be used in design observation.