| MEMS-based wearable motion sensors(accelerometers,gyroscopes,magnetometers)have been widely used in human motion analysis,such as gait analysis,fall risk assessment,gesture recognition,for their advantages of small size,low cost and ease of use.Motion sensors are not only used to collect raw acceleration and angular velocity but also used to estimate the orientation and displacement of body segments.Accurate orientation and displacement estimation are critical in human motion analysis.However,the estimation accuracy of orientation and displacement is unsatisfactory and easily influenced by magnetic disturbance and acceleration distortion.These problems have become a bottleneck of their extensive applications in motion analysis.This dissertation focuses on accurate orientation and displacement estimation based on motion sensors.We have carried out the following work:1)We performed a comparison study on orientation estimation algorithms that focus on dealing with magnetic disturbance.We divided the orientation estimation into four components.Then,we systematically reviewed the commonly used methods of each component and their influences on the estimation of attitude and heading.A new standardized testing procedure has been developed to objectively assess the performance of each algorithm against magnetic disturbance.Finally,we compared four representative algorithms using the proposed testing procedure.Based upon the results,the strength and weakness of the existing algorithms were easily examined,and suggestions were presented for selecting a proper or developing new orientation estimation algorithms.2)We proposed a gradient descent based adaptive orientation estimation algorithm in the presence of magnetic disturbance.The algorithm adaptively modulates the parameters according to the motion state of the sensor and the severity of magnetic disturbance,thus the orientation estimation is immune to magnetic disturbance in static state and has smaller error in the presence of magnetic disturbance in dynamic state.3)Besides magnetic disturbance,the acceleration distortion is another main factor that influences the accuracy of orientation estimation,and it exists with the motion of the sensor,which should also be considered in orientation estimation.We first implemented a quaternion-based two-step complementary filter as the base sensor fusion algorithm.Then,we proposed a novel finite state machine(FSM)based adaptive strategy.The experimental results demonstrate that the proposed sensor fusion method performed well against external acceleration and magnetic disturbance compared with existing methods.Especially,the proposed method showed very high accuracy in a 60-second continuous combined distorted condition,where the RMSEs of the roll,pitch and yaw were 0.63°,0.83°,and 0.96°,respectively.4)We applied the FSM based sensor fusion algorithm to spatial gait parameters estimation in gait analysis.We also analyzed other factors that influence the accuracy of foot displacement estimation,and then proposed a set of new procedures improving each component in foot displacement estimation.With the proposed method,the accuracy(mean ±SD)of maximal heel clearance and minimal toe clearance were-0.42±0.25 cm and 0.13±0.24cm,the stride length was 0.39±0.37 cm.The estimation accuracy is significantly improved compared with other studies.Most importantly,the proposed method does not require any post-correction and flat-floor assumption,which is more applicable and practical.This dissertation provides robust orientation estimation method in the presence of magnetic and acceleration distortions and displacement estimation method for gait parameters estimation.Moreover,it also enables one to gain insight into motion sensor-based orientation and displacement estimation.The presented accurate and practical orientation and displacement estimation methods will promote the application of wearable motion sensors. |
PDF Full Text Request