Development of a Low-Cost and Portable Walker-Based Human Motion Estimation System

Accurate human posture estimations are important for patients' motor function assessment at home rehabilitations. Current high-precision posture detection devices are expensive and not portable for use at home or in the community. In this work, a walker-based human motion estimation (WHUME) system is developed to achieve low-cost three-dimensional human posture estimation using only one monocular camera and four IMUs, and these sensors are integrated into the walker platform, which effectively improves the portability of the device. To achieve real-time posture estimation based on the WHUME system, a stepwise iterative optimization algorithm is developed to efficiently fuse the sparse-sensor information and uncertain priori information from the whole body, whose estimation accuracy can be higher than that of approaches using more sensors. Based on the iterative estimations of limb lengths, no precise lengths are required in advance in the algorithm, which improves the convenience of use. Spatial correlation constraints used in the algorithm guarantee the rationality of estimated limb positions. Meanwhile, results of the posture estimations based on the WHUME system can be used to obtain high-accuracy gait parameter recognitions.