Reliable full-body tracking in confined indoor spaces depends less on a single sensor and more on combining complementary modalities to overcome occlusion, clutter, and privacy constraints. Evidence from established research shows certain sensors excel in specific roles: depth cameras capture geometry, inertial sensors capture limb dynamics, RGB pose models infer joints from appearance, and radio-frequency devices sense through occlusions. Choosing and fusing these reduces failure modes while balancing cost, latency, and user acceptance.
Depth and structured-light sensors
Depth sensors provide direct 3D geometry that simplifies skeleton fitting. Jamie Shotton Microsoft Research Cambridge demonstrated the value of depth-based pose recognition for real-time skeletal tracking, showing robustness to clothing and lighting changes compared with pure RGB. Depth cameras such as time-of-flight and structured-light devices are effective in small rooms because they yield accurate distances at short ranges, but they struggle with reflective surfaces and require line-of-sight, which increases sensitivity to furniture and crowding.Inertial measurement units and wearable IMUs
Body-worn inertial measurement units provide continuous orientation and acceleration data even when limbs are occluded. Huub Luinge and Jan Veltink University of Twente have published foundational work on inertial-based orientation estimation and biomechanical use. Commercial systems from Xsens combine multiple IMUs into a body model to achieve drift-corrected full-body capture indoors. Wearables mitigate occlusion but introduce attachment burden and potential cultural resistance to wearing sensors in shared domestic spaces.Vision-based RGB pose estimation
2D and 3D pose estimators from RGB imagery complement depth and IMUs when camera coverage is limited. Zhe Cao Carnegie Mellon University developed real-time multi-person pose methods that can run on single cameras to extract joints from silhouettes and part affinities. RGB-based models are sensitive to lighting and require consent for video capture, raising privacy concerns in bedrooms or communal housing.Radio-frequency and wireless sensing
Radio-frequency techniques using WiFi channel state information or mmWave radar can detect body movements without cameras. Fadel Adib Massachusetts Institute of Technology has shown RF approaches that penetrate occluders and preserve some privacy because they do not capture identifiable images. RF sensing can be influenced by building layout and multipath effects, so accuracy varies with environment.Combining modalities through sensor fusion addresses causes of failure—occlusion, drift, and environmental interference—while enabling trade-offs between intrusiveness and accuracy. In small indoor spaces the practical consequence is that hybrid systems, for example depth plus IMUs or RGB plus RF, deliver the most reliable full-body tracking while requiring careful attention to privacy, cultural acceptability, and room-specific calibration.