Vladimir Joukov

Human motion segmentation by data point classification

Contemporary physiotherapy and rehabilitation practice uses subjective measures for motion evaluation and requires time-consuming supervision. Algorithms that can accurately segment patient movement would provide valuable data for progress tracking and on-line patient feedback. In this paper, we propose a two-class classifier approach to label each data point in the patient movement data as either a segment point or a non-segment point. The proposed technique was applied to 20 healthy subjects performing lower body rehabilitation exercises, and achieves a segmentation accuracy of 82%.

Online tracking of the lower body joint angles using IMUs for gait rehabilitation

An important field in physiotherapy is the rehabilitation of gait. A continuous assessment and progress tracking of a patients ability to walk is of clinical interest. Unfortunately the tools available to the therapists are very time-consuming and subjective. Non-intrusive, small, wearable, wireless sensors can be worn by the patients and provide inertial measurements to estimate the pose of the lower body during walking. For this purpose, we propose two different kinematic models of the human lower body. We use an Extended Kalman Filter to estimate the joint angles and show that a variety of sensors, such as accelerometers, gyroscopes, and motion capture markers, can be used and fused together to aid the joint angle estimate. The algorithm is validated on gait data collected from healthy participants.

A constrained Extended Kalman Filter for dynamically consistent inverse kinematics and inertial parameters identification

This paper presents a method for the real-time determination of joint angles, velocities, accelerations and joint torques of a human. The proposed method is based on a constrained Extended Kalman Filter that combines stereophotogrammetric and dynamometric data. In addition to the joint variables, subject-specific segment lengths and inertial parameters are identified. Constraints are added to the filter, by restricting the optimal Kalman gain, in order to obtain physically consistent parameters. An optimal tuning procedure of the filters gains and a sensitivity analysis is presented. The method is validated in the plane on four human subjects and shows very good tracking of skin markers with a RMS difference lower than 15 mm. External ground reaction forces and resultant moment are also accurately estimated with an RMS difference below 3 N and 6 N.m, respectively.

Monitoring of Hip and Knee Joint Angles Using a Single Inertial Measurement Unit During Lower Limb Rehabilitation

This paper investigated the possibility of estimating 3D lower limb joint kinematics during five popular rehabilitation exercises of the hip and knee joints based on the data collected from a single inertial measurement unit located on the shank. The leg was modeled as a four-degree-of-freedom serial chain, and the relevant joint angles were represented by Fourier series. A least square approach based on the minimization of the difference between the measured and estimated 3D linear accelerations and angular velocities was used to solve the related analytical problem. The approach was validated on ten healthy young volunteers (ten trials each), comparing the proposed approach with the measurements collected through a stereophotogrammetric system. The average root mean square differences between the estimated joint angles and those reconstructed with the stereophotogrammetric system were inferior than 3.2° with correlation coefficients higher than 0.85.

A constrained extended Kalman filter for the optimal estimate of kinematics and kinetics of a sagittal symmetric exercise

This paper presents a method for real-time estimation of the kinematics and kinetics of a human body performing a sagittal symmetric motor task, which would minimize the impact of the stereophotogrammetric soft tissue artefacts (STA). The method is based on a bi-dimensional mechanical model of the locomotor apparatus the state variables of which (joint angles, velocities and accelerations, and the segments lengths and inertial parameters) are estimated by a constrained extended Kalman filter (CEKF) that fuses input information made of both stereophotogrammetric and dynamometric measurement data. Filter gains are made to saturate in order to obtain plausible state variables and the measurement covariance matrix of the filter accounts for the expected STA maximal amplitudes. We hypothesised that the ensemble of constraints and input redundant information would allow the method to attenuate the STA propagation to the end results. The method was evaluated in ten human subjects performing a squat exercise. The CEKF estimated and measured skin marker trajectories exhibited a RMS difference lower than 4mm, thus in the range of STAs. The RMS differences between the measured ground reaction force and moment and those estimated using the proposed method (9N and 10Nm) were much lower than obtained using a classical inverse dynamics approach (22N and 30Nm). From the latter results it may be inferred that the presented method allows for a significant improvement of the accuracy with which kinematic variables and relevant time derivatives, model parameters and, therefore, intersegmental moments are estimated.

Rhythmic EKF for pose estimation during gait

Accurate estimation of lower body pose during gait is useful in a wide variety of applications, including design of bipedal walking strategies, active prosthetics, exoskeletons, and physical rehabilitation. In this paper an algorithm is developed to estimate joint kinematics during rhythmic motion such as walking, using inertial measurement units attached at the waist, knees, and ankles. The proposed approach combines the extended Kalman filter with a canonical dynamical system to estimate joint angles, positions, and velocities for 3 dimensional rhythmic lower body movement. The system incrementally learns the rhythmic motion over time, improving the estimate over a regular extended Kalman filter, and segmenting the motion into repetitions. The algorithm is validated in simulation and on real human walking data. It is shown to improve joint acceleration and velocity estimates over regular extended Kalman Filter by 40% and 37% respectively.

Full-body multi-primitive segmentation using classifiers

During human-robot interaction, the robot observes a continuous stream of time-series data capturing the behaviour of the human and any changes in the environment. For applications such as imitation learning, intention and gesture recognition, the time-series data is typically segmented into action or motion primitives, requiring accurate and online temporal segmentation. This paper casts the time-series segmentation problem into a two-class classification problem, labelling each data point as either a segment edge or a within-segment point, and applies several common classifiers to a set of full body motion data. The support vector machine combined with principal component analysis dimensionality reduction were found to perform best, with a classification F1 score of 91% when applied to novel exemplars. The proposed approach can also generalize to motions unseen during training, achieving a classification F1 score of 83% when applied to novel motions.

Full body human motion estimation on lie groups using 3D marker position measurements

This paper proposes a new algorithm for full body human motion estimation using 3D marker position measurements. The joints are represented with Lie group members, including special orthogonal groups SO(2) and SO(3), and a special euclidean group SE(3). We employ the Lie Group Extended Kalman Filter (LG-EKF) for stochastic inference on groups, thus explicitly accounting for the non-euclidean geometry of the state space, and provide the derivation of the LG-EKF recursion for articulated motion estimation. We evaluate the performance of the proposed algorithm in both simulation and on real-world motion capture data, comparing it with the Euler angles based EKF. The results show that the proposed filter significantly outperforms the Euler angles based EKF, since it estimates human motion more accurately and is not affected by gimbal lock.

Constrained dynamic parameter estimation using the Extended Kalman Filter

In this paper we present a real-time method for identification of the dynamic parameters of a manipulator and its load using kinematic measurements and either joint torques or force and moment at the base. The parameters are estimated using the Extended Kalman Filter and constraints are imposed using Sigmoid functions to ensure the parameters remain within their physically feasible ranges, such as links having positive masses and moments of inertia. Identified parameters can be used in model based controllers. The presented approach is validated through simulation and on data collected with the Barret WAM manipulator. Using the estimated parameters instead of ones provided by the manufacturer greatly improves joint torque prediction.

Rhythmic Extended Kalman Filter for Gait Rehabilitation Motion Estimation and Segmentation

This paper proposes a method to enable the use of non-intrusive, small, wearable, and wireless sensors to estimate the pose of the lower body during gait and other periodic motions and to extract objective performance measures useful for physiotherapy. The Rhythmic Extended Kalman Filter (Rhythmic-EKF) algorithm is developed to estimate the pose, learn an individualized model of periodic movement over time, and use the learned model to improve pose estimation. The proposed approach learns a canonical dynamical system model of the movement during online observation, which is used to accurately model the acceleration during pose estimation. The canonical dynamical system models the motion as a periodic signal. The estimated phase and frequency of the motion also allow the proposed approach to segment the motion into repetitions and extract useful features, such as gait symmetry, step length, and mean joint movement and variance. The algorithm is shown to outperform the extended Kalman filter in simulation, on healthy participant data, and stroke patient data. For the healthy participant marching dataset, the Rhythmic-EKF improves joint acceleration and velocity estimates over regular EKF by 40% and 37%, respectively, estimates joint angles with 2.4° root mean squared error, and segments the motion into repetitions with 96% accuracy.