Statistical method for prediction of gait kinematics with Gaussian process regression

We propose a novel methodology for predicting human gait pattern kinematics based on a statistical and stochastic approach using a method called Gaussian process regression (GPR). We selected 14 body parameters that significantly affect the gait pattern and 14 joint motions that represent gait kinematics. The body parameter and gait kinematics data were recorded from 113 subjects by anthropometric measurements and a motion capture system. We generated a regression model with GPR for gait pattern prediction and built a stochastic function mapping from body parameters to gait kinematics based on the database and GPR, and validated the model with a cross validation method. The function can not only produce trajectories for the joint motions associated with gait kinematics, but can also estimate the associated uncertainties. Our approach results in a novel, low-cost and subject-specific method for predicting gait kinematics with only the subject's body parameters as the necessary input, and also enables a comprehensive understanding of the correlation and uncertainty between body parameters and gait kinematics.     

A kinematic model of the human hand to evaluate its prehensile capabilities.

A kinematic model has been developed for simulation and prediction of the prehensile capabilities of the human hand. The kinematic skeleton of the hand is characterized by ideal joints and simple segments. Finger-joint angulation is characterized by yaw (abduction-adduction), pitch (flexion-extension) and roll (axial rotation) angles. The model is based on an algorithm that determines contact between two ellipsoids, which are used to approximate the geometry of the cutaneous surface of the hand segments. The model predicts the hand posture (joint angles) for power grasp of ellipsoidal objects by 'wrapping' the fingers around the object. Algorithms for two grip types are included: (1) a transverse volar grasp, which has the thumb abducted for added power; and (2) a diagonal volar grasp, which has the thumb adducted for an element of precision. Coefficients for estimating anthropometric parameters from hand length and breadth are incorporated in the model. Graphics procedures are included for visual display of the model. In an effort to validate the predictive capabilities of the model, joint angles were measured on six subjects grasping circular cylinders of various diameters and these measured joint angles were compared with angles predicted by the model. Sensitivity of the model to the various input parameters was also determined. On an average, the model predicted joint flexion angles that were 5.3% or 2.8 degrees +/- 12.2 degrees larger than the measured angles. Good agreement was found for the MCP and PIP joints, but results for DIP were more variable because of its dependence on the predictions for the proximal joints.     

Influence of body segments' parameters estimation models on inverse dynamics solutions during gait

The purpose of the present study was to examine the influence of anthropometric data on joint kinetics during gait. We particularly focused on the sensitivity of inverse dynamics solutions to the use of models for body segment parameters (BSP) estimation. Six often used estimation models were selected to provide BSP values for the three segments of the lower limb. Kinematics and dynamics were sampled from seven subjects performing barefoot gait at three different speeds. Joint kinetics were estimated with the bottom-up method using BSP values derived from each estimation model as anthropometric inputs. The BSP estimates were highly sensitive to the model used with deviations ranging from at least 9.73% up to 60%. Maximal variations of peak values for the hip joint flexion/extension moment during the swing phase were 20.11%. Hence, our findings suggest that the influence of BSP cannot be neglected. Observed deviations are especially due to the effect of varying simultaneously the mass, moments of inertia and the center of mass location values, according to the underlying relationship of interdependency linking each component. Considering both the differences found in joint kinetics and the level of accuracy of BSP models, evidence is provided that using multiple regression BSP estimation functions derived from Zatsiorsky and Seluyanov should be recommended to assess joint kinetics.     

Validity of time series kinematical data as measured by a markerless motion capture system on a flatland for gait assessment

As a cost-effective, clinician-friendly gait assessment tool, the Kinect v2 sensor may be effective for assessing lower extremity joint kinematics. This study aims to examine the validity of time series kinematical data as measured by the Kinect v2 on a flatland for gait assessment. In this study, 51 healthy subjects walked on a flatland while kinematic data were extracted concurrently using the Kinect and Vicon systems. The kinematic outcomes comprised the hip and knee joint angles. Parallel translation of Kinect data obtained throughout the gait cycle was performed to minimize the differences between the Kinect and Vicon data. The ensemble curves of the hip and knee joint angles were compared to investigate whether the Kinect sensor can consistently and accurately assess lower extremity joint motion throughout the gait cycle. Relative consistency was assessed using Pearson correlation coefficients. Joint angles measured by the Kinect v2 followed the trend of the trajectories made by the Vicon data in both the hip and knee joints in the sagittal plane. The trajectories of the hip and knee joint angles in the frontal plane differed between the Kinect and Vicon data. We observed moderate to high correlation coefficients of 20%–60% of the gait cycle, and the largest difference between Kinect and Vicon data was 4.2°. Kinect v2 time series kinematical data obtained on the flatland are validated if the appropriate correction procedures are performed. Future studies are warranted to examine the reproducibility and systematic bias of the Kinect v2.     

Kinematic Analysis and Experimental Verification on the Locomotion of Gecko

This paper presents a kinematic analysis of the locomotion of a gecko,and experimental verification of the kinematicmodel.Kinematic analysis is important for parameter design,dynamic analysis,and optimization in biomimetic robot research.The proposed kinematic analysis can simulate,without iteration,the locomotion of gecko satisfying the constraint conditionsthat maintain the position of the contacted feet on the surface.So the method has an advantage for analyzing the climbing motionof the quadruped mechanism in a real time application.The kinematic model of a gecko consists of four legs based on 7-degreesof freedom spherical-revolute-spherical joints and two revolute joints in the waist.The motion of the kinematic model issimulated based on measurement data of each joint.The motion of the kinematic model simulates the investigated real gecko’smotion by using the experimental results.The analysis solves the forward kinematics by considering the model as a combinationof closed and open serial mechanisms under the condition that maintains the contact positions of the attached feet on the ground.The motions of each joint are validated by comparing with the experimental results.In addition to the measured gait,three othergaits are simulated based on the kinematic model.The maximum strides of each gait are calculated by workspace analysis.Theresult can be used in biomimetic robot design and motion planning.     

Kinematic correlates of walking cadence in the foot

Evidence has frequently been reported of modifications in gait patterns within the lower limb related to the cadence of walking. Most reports have concerned relationships between cadence and kinematic and the kinetic changes occurring in the main joints and muscles of the lower limb as a whole. The aim of the present study was to assess whether significant changes are also measurable in kinematics of the foot segments. An existing 15 marker-set protocol allowed a four-segment foot and shank model to be defined for relative rotations between the segments to be calculated. Stereophotogrammetry was employed to record marker position data from ten subjects walking at three cadences. The slow- and normal cadence datasets showed similar profiles of joint rotation in three anatomical planes, but significant differences were found between these and the fast cadence. At all joints, frame-by-frame statistical analysis revealed increased dorsiflexion from heel-strike to midstance (p<0.05) and increased plantarflexion from midstance to toe-off (p<0.05) with increasing cadence. From foot-flat to heel-rise, the fast cadence kinematic data showed a decreased range of motion in the sagittal-plane between forefoot and rearfoot (3.2°±1.2° at slow cadence; 2.0°±0.8° at fast cadence; p<0.05). The cadences imposed and the multisegment protocol revealed significant kinematic changes in the joints of the foot during barefoot walking.     

Dynamic analysis of load carriage biomechanics during level walking

This paper describes an investigation into the biomechanical effects of load carriage dynamics on human locomotion performance. A whole body, inverse dynamics gait model has been developed which uses only kinematic input data to define the gait cycle. To provide input data, three-dimensional gait measurements have been conducted to capture whole body motion while carrying a backpack. A nonlinear suspension model is employed to describe the backpack dynamics. The model parameters for a particular backpack system can be identified using a dynamic load carriage test-rig. Biomechanical assessments have been conducted based on combined gait and pack simulations. It was found that the backpack suspension stiffness and damping have little effect on human locomotion energetics. However, decreasing suspension stiffness offers important biomechanical advantages. The peak values of vertical pack force, acting on the trunk, and lower limb joint loads are all moderated. This would reduce shoulder strap pressures and the risk of injury when heavy loads are carried.     

Upper extremity kinematic and kinetic adaptations during a fatiguing repetitive task

Repetitive low-force contractions are common in the workplace and yet can lead to muscle fatigue and work-related musculoskeletal disorders. The current study aimed to investigate potential motion adaptations during a simulated repetitive light assembly work task designed to fatigue the shoulder region, focusing on changes over time and age-related group differences. Ten younger and ten older participants performed four 20-min task sessions separated by short breaks. Mean and variability of joint angles and scapular elevation, joint net moments for the shoulder, elbow, and wrist were calculated from upper extremity kinematics recorded by a motion tracking system. Results showed that joint angle and joint torque decreased across sessions and across multiple joints and segments. Increased kinematic variability over time was observed in the shoulder joint; however, decreased kinematic variability over time was seen in the more distal part of the upper limb. The changes of motion adaptations were sensitive to the task-break schedule. The results suggested that kinematic and kinetic adaptations occurred to reduce the biomechanical loading on the fatigued shoulder region. In addition, the kinematic and kinetic responses at the elbow and wrist joints also changed, possibly to compensate for the increased variability caused by the shoulder joint while still maintaining task requirements. These motion strategies in responses to muscle fatigue were similar between two age groups although the older group showed more effort in adaptation than the younger in terms of magnitude and affected body parts.     

Gait posture estimation using wearable acceleration and gyro sensors

A method for gait analysis using wearable acceleration sensors and gyro sensors is proposed in this work. The volunteers wore sensor units that included a tri-axis acceleration sensor and three single axis gyro sensors. The angular velocity data measured by the gyro sensors were used to estimate the translational acceleration in the gait analysis. The translational acceleration was then subtracted from the acceleration sensor measurements to obtain the gravitational acceleration, giving the orientation of the lower limb segments. Segment orientation along with body measurements were used to obtain the positions of hip, knee, and ankle joints to create stick figure models of the volunteers. This method can measure the three-dimensional positions of joint centers of the hip, knee, and ankle during movement. Experiments were carried out on the normal gait of three healthy volunteers. As a result, the flexion–extension (F–E) and the adduction–abduction (A–A) joint angles of the hips and the flexion–extension (F–E) joint angles of the knees were calculated and compared with a camera motion capture system. The correlation coefficients were above 0.88 for the hip F–E, higher than 0.72 for the hip A–A, better than 0.92 for the knee F–E. A moving stick figure model of each volunteer was created to visually confirm the walking posture. Further, the knee and ankle joint trajectories in the horizontal plane showed that the left and right legs were bilaterally symmetric.     

Influence of the 3D inverse dynamic method on the joint forces and moments during gait

The joint forces and moments are commonly used in gait analysis. They can be computed by four different 3D inverse dynamic methods proposed in the literature, either based on vectors and Euler angles, wrenches and quaternions, homogeneous matrices, or generalized coordinates and forces. In order to analyze the influence of the inverse dynamic method, the joint forces and moments were computed during gait on nine healthy subjects. A ratio was computed between the relative dispersions (due to the method) and the absolute amplitudes of the gait curves. The influence of the inverse dynamic method was negligible at the ankle (2%) but major at the knee and the hip joints (40%). This influence seems to be due to the dynamic computation rather than the kinematic computation. Compared to the influence of the joint center location, the body segment inertial parameter estimation, and more, the influence of the inverse dynamic method is at least of equivalent importance. This point should be confirmed with other subjects, possibly pathologic, and other movements.     

Evaluation of lower limb cross planar kinetic connectivity signatures post-stroke

Following stroke, aberrant three dimensional multijoint gait impairments emerge that present in kinematic asymmetries such as circumduction. A precise pattern of cross-planar coordination may underlie abnormal hemiparetic gait as several studies have underscored distinctive neural couplings between medio-lateral control and sagittal plane progression during walking. Here we investigate potential neuromechanical constraints governing abnormal multijoint coordination post-stroke. 15 chronic monohemispheric stroke patients and 10 healthy subjects were recruited. Coupled torque production patterns were assessed using a volitional isometric torque generation task where subjects matched torque targets for a primary joint in 4 directions while receiving visual feedback of the magnitude and direction of the torque. Secondary torques at other lower limb joints were recorded without subject feedback. We find that common features of cross-planar connectivity in stroke subjects include statistically significant frontal to sagittal plane kinetic coupling that overlay a common sagittal plane coupling in healthy subjects. Such coupling is independent of proximal or distal joint control and limb biomechanics. Principal component analysis of the stroke aggregate kinetic signature reveals unique abnormal frontal plane coupling features that explain a larger percentage of the total torque coupling variance. This study supports the idea that coupled cross-planar kinetic outflow between the lower limb joints uniquely emerges during pathological control of frontal plane degrees of freedom resulting in a generalized extension of the limb. It remains to be seen if a pattern of lower limb motor outflow that is centrally mediated contributes to abnormal hemiparetic gait.     

Double-step registration of in vivo stereophotogrammetry with both in vitro 6-DOFs electrogoniometry and CT medical imaging

Standard registration techniques of bone morphology to motion analysis data often lead to unsatisfactory motion simulation because of discrepancies during the location of anatomical landmarks in the datasets. This paper describes an iterative registration method of a three-dimensional (3D) skeletal model with both 6 degrees-of-freedom joint kinematics and standard motion analysis data. The method is demonstrated in this paper on the lower limb. The method includes two steps. A primary registration allowed synchronization of in vitro kinematics of the knee and ankle joints using flexion/extension angles from in vivo gait analysis. Results from primary registration were then improved by a so-called advanced registration, which integrated external constraints obtained from experimental gait pre-knowledge. One cadaver specimen was analyzed to obtain both joint kinematics of knee and ankle joints using 3D electrogoniometry, and 3D bone morphology from medical imaging data. These data were registered with motion analysis data from a volunteer during the execution of locomotor tasks. Computer graphics output was implemented to visualize the results for a motion of sitting on a chair. Final registration results allowed the observation of both in vivo motion data and joint kinematics from the synchronized specimen data. The method improved interpretation of gait analysis data, thanks to the combination of realistic 3D bone models and joint mechanism. This method should be of interest both for research in gait analysis and medical education. Validation of the overall method was performed using RMS of the differences between bone poses estimated after registration and original data from motion analysis.     

Quantification of gait parameters with inertial sensors and inverse kinematics

Measuring human gait is important in medicine to obtain outcome parameter for therapy, for instance in Parkinson’s disease. Recently, small inertial sensors became available which allow for the registration of limb-position outside of the limited space of gait laboratories. The computation of gait parameters based on such recordings has been the subject of many scientific papers. We want to add to this knowledge by presenting a 4-segment leg model which is based on inverse kinematic and Kalman filtering of data from inertial sensors. To evaluate the model, data from four leg segments (shanks and thighs) were recorded synchronously with accelerometers and gyroscopes and a 3D motion capture system while subjects (n = 12) walked at three different velocities on a treadmill. Angular position of leg segments was computed from accelerometers and gyroscopes by Kalman filtering and compared to data from the motion capture system. The four-segment leg model takes the stance foot as a pivotal point and computes the position of the remaining segments as a kinematic chain (inverse kinematics). Second, we evaluated the contribution of pelvic movements to the model and evaluated a five segment model (shanks, thighs and pelvis) against ground-truth data from the motion capture system and the path of the treadmill.ResultsWe found the precision of the Kalman filtered angular position is in the range of 2–6° (RMS error). The 4-segment leg model computed stride length and length of gait path with a constant undershoot of 3% for slow and 7% for fast gait. The integration of a 5th segment (pelvis) into the model increased its precision. The advantages of this model and ideas for further improvements are discussed.     

On the estimation of joint kinematics during gait.

In gait analysis, the concepts of Euler and helical (screw) angles are used to define the three-dimensional relative joint angular motion of lower extremities. Reliable estimation of joint angular motion depends on the accurate definition and construction of embedded axes within each body segment. In this paper, using sensitivity analysis, we quantify the effects of uncertainties in the definition and construction of embedded axes on the estimation of joint angular motion during gait. Using representative hip and knee motion data from normal subjects and cerebral palsy patients, the flexion-extension axis is analytically perturbed +/- 15 degrees in 5 degrees steps from a reference position, and the joint angles are recomputed for both Euler and helical angle definitions. For the Euler model, hip and knee flexion angles are relatively unaffected while the ab/adduction and rotation angles are significantly affected throughout the gait cycle. An error of 15 degrees in the definition of flexion-extension axis gives rise to maximum errors of 8 and 12 degrees for the ab/adduction angle, and 10-15 degrees for the rotation angles at the hip and knee, respectively. Furthermore, the magnitude of errors in ab/adduction and rotation angles are a function of the flexion angle. The errors for the ab/adduction angles increase with increasing flexion angle and for the rotation angle, decrease with increasing flexion angle. In cerebral palsy patients with flexed knee pattern of gait, this will result in distorted estimation of ab/adduction and rotation. For the helical model, similar results are obtained for the helical angle and associated direction cosines.(ABSTRACT TRUNCATED AT 250 WORDS)     

A three-dimensional kinematic and dynamic model of the lower limb

A model describing the kinematics and dynamics of the lower limb is presented. The lower limb is modeled as a sequence of four rigid links connected by three universal rotary joints representing the hip, knee and ankle joints. Each joint is modeled as a sequence of three single axis rotational joints thus ascribing to the lower limb a total of 12 degrees of freedom. A method is described to measure the gait variables so that all nine angles can be computed based on the positions of nine markers placed on the subject during a gait study. The gait variables are then used in an iterative Newton-Euler formulation to compute the moments exerted about the axes of each joint during gait.     

In vivo determination of the anatomical axes of the ankle joint complex: An optimization approach

This study investigates the feasibility of a subject-specific three-dimensional model of the ankle joint complex for kinematic and dynamic analysis of movement. The ankle joint complex was modelled as a three-segment system, connected by two ideal highe joints: the talocrural and the subtalar joint. A mathematical formulation was developed to express the three-dimensional translation and rotation between the foot and shank segments as a function of the two joint angles, and 12 model parameters describing the locations of the joint axes. An optimization method was used to fit the model parameters to three-dimensional kinematic data of foot and shank markers, obtained during test movements throughout the entire physiological range of motion of the ankle joint. The movement of the talus segment, which cannot be measured non-invasively, is not necessary for the analysis.

This optimization method was used to determine the position and orientation of the joint axes in 14 normal subjects. After optimization, the discrepancy between the best fitting model and actual marker kinematics was between 1 and 3 mm for all subjects. The predicted inclination of the subtalar joint axis from the horizontal plane was 37.4±2.7°, and the medial deviation was 18.0±16.2°. The lateral side of the talucrural axis was directed slightly posteriorly (6.8±8.1°), and inclined downward by 7.0±5.4°. These results are similar to previously reported typical results from anatomical, in vitro, studies. Reproducibility was evaluated by repeated testing of one subject, which resulted in variations of about one-fifth of the standard deviation within the group, the inclination of the subtalar joint axis was significantly correlated to the arch height and a radiographic ‘tarsal index’. It is concluded that this optimization method provides the opportunity to incorporate inter-individual anatomical differences into kinematic and dynamic analysis of the ankle joint complex. This allows a more functional interpretation of kinematic data, and more realistic estimates of internal forces.     

Determination of patient-specific multi-joint kinematic models through two-level optimization

Dynamic patient-specific musculoskeletal models have great potential for addressing clinical problems in orthopedics and rehabilitation. However, their predictive capability is limited by how well the underlying kinematic model matches the patient's structure. This study presents a general two-level optimization procedure for tuning any multi-joint kinematic model to a patient's experimental movement data. An outer level optimization modifies the model's parameters (joint position and orientations) while repeated inner level optimizations modify the model's degrees of freedom given the current parameters, with the goal of minimizing errors between model and experimental marker trajectories. The approach is demonstrated by fitting a 27 parameter, three-dimensional, 12 degree-of-freedom lower-extremity kinematic model to synthetic and experimental movement data for isolated joint (hip, knee, and ankle) and gait (full leg) motions. For noiseless synthetic data, the approach successfully recovered the known joint parameters to within an arbitrarily tight tolerance. When noise was added to the synthetic data, root-mean-square (RMS) errors between known and recovered joint parameters were within 10.4 degrees and 10 mm. For experimental data, RMS marker distance errors were reduced by up to 62% compared to methods that estimate joint parameters from anatomical landmarks. Optimized joint parameters found using a loaded full-leg gait motion differed significantly from those found using unloaded individual joint motions. In the future, this approach may facilitate the creation of dynamic patient-specific musculoskeletal models for predictive clinical applications.     

Quantitative analysis of upper limbs during gait: a marker set protocol

Purpose: to develop a marker set for simultaneously assessing upper and lower limb biomechanics during gait.Methods: 24 healthy young subjects (mean age: 23.80 years) were assessed quantitatively using an optoelectronic system, two force platform and a video system. Passive markers were positioned according to the proposed marker set which enables acquiring the upper and lower limb movement simultaneously during Gait Analysis. In addition to the traditional parameters obtained from Gait Analysis, the shoulder and elbow angles were computed from markers coordinates of upper limbs; then, some significant parameters were identified and calculated. From shoulder and elbow position, angles, angular velocities, angular acceleration, moments, and powers were calculated for shoulder and elbow joints. Results: Kinematic and kinetic data were obtained in the three planes (sagittal, frontal, and transversal) for the shoulder and in the sagittal plane for the elbow. Normative ranges were obtained for these parameters from data of healthy participants. Conclusions: The proposed experimental set-up enables simultaneous assessment of upper and lower limb movement during gait. Thus, no further trials are required in addition to those acquired during standard gait analysis in order to assess upper limb motion, which also makes the experimental set-up feasible for clinical applications.