Spatial analysis and evaluation are becoming increasingly common as new technologies enable users, designers, and researchers to study spatial motion patterns without relying on manual notations for observations. While ideas related to motion and space have been studied in other fields such as industrial engineering, choreography, and computer science, the understanding of efficiency and quality in architectural spaces through motion has not been widely explored. This research applies techniques in computer vision to analyse human body motion in architectural spaces as a measure of experience and engagement. A taxonomy framework is proposed to categorize human motion components relevant to spatial interactions, for analysis through computer vision. A technical case study developed upon a machinelearning-aided model is used to test a selection of the proposed framework within domestic kitchen environments. This contribution adds further perspective to wider research explorations in the quality, inclusivity, engagement, and efficiency of architectural spaces through computer-aided tools.