Periodic Gait Research Articles

LFN: Based on the Convolutional Neural Network of Gait Recognition Method

In this paper, a novel method based on convolutional neural network (CNN) to address gait recognition was proposed. Gait is a unique biologic feature that the feature almost cannot be altered. Existing gait recognition virtually based on a traditional method such as Gait Energy Image (GEI). GEI has various gait silhouette sequence images and put this image together. Meanwhile, those images must in the same gait period, which lacks of flexibility. To address this issue, a different gait recognition model, called LFN, based on a convolution neural network (CNN) was proposed. The model is composed of three convolution layers in parallel. It can be trained in an end-to-end manner and each sample gait silhouette in the uniform gait period does not to be required. Our working base on the OU-ISIR large population gait dataset. Being based on the result of the experiments, our network can learn the significant features of each sample gait silhouette sequence effectively, at the same time, our model can achieve high and stable accuracy in three experiments.

Modification‐based periodic gait planning for a 3‐D bipedal robot with two underactuated DOFs

AbstractThis work considers the periodic gait planning of a 3‐D bipedal robot with two underactuated DOFs. Instead of planning a periodic gait by a traditional direct parameter optimization search, which is usually time‐consuming, we present a novel modification‐based periodic motion planning method, which consists of two steps. Firstly, an initial gait is obtained by relaxing the periodicity constraints in the optimization search. Then, the initial gait is modified to be a periodic one. To achieve a successful gait modification, the gait modification theory is firstly presented, and a gait modification strategy is then proposed. Finally, the presented method is illustrated on the underactuated 3‐D biped, and a comparative simulation study is provided. It is shown that, compared with the traditional method, the time cost in the presented method is significantly reduced.

A Comparison of the Center of Pressure Trajectory Between Dynamic Response Prosthetic Feet Having Different Ankle Designs

ABSTRACT Introduction Commercially available dynamic response (DR) prosthetic feet have unique ankle designs, which provide different degrees of ankle motion. Differences in ankle motion between the prosthetic and anatomical foot creates an asymmetry in the locomotor system and affects its functions of shock absorption, progression, and stabilization. Ankle motion during gait can be assessed by quantifying the center of pressure (CoP) movement during the stance period of gait. Although previous studies have compared gait parameters between DR and non-DR feet, no studies have examined difference in CoP trajectories among common designs of DR feet. Methods LP-Variflex, Talux, and Variflex prosthetic feet were used as test feet to represent the common DR feet, that is, low-profile (LP), multi-axial (MX), and J-shaped (JS) ankle designs, respectively. Five subjects with unilateral transtibial amputation (TTA), six subjects with unilateral transfemoral amputation (TFA), and three healthy controls participated in the study. Subjects were prosthetic users for at least 18 months and were classified as K-level 3 or K-level 4 functional level. Each subject received standardized functional prosthetic gait training and accommodation with each test foot. For data collection, subjects walked on level ground using standardized shoes and CoP data were collected using in-sole sensors. The testing order of feet was randomized. The A-P and M-L location of CoP was determined at the following gait phases: initial contact (2% of gait cycle), early midstance (15% of gait cycle), terminal stance (35% of gait cycle), preswing (50% of gait cycle), and before toe-off (60% of gait cycle). The maximum A-P and M-L CoP excursion was also calculated. The A-P displacement of the CoP was normalized to the foot length, and the M-L displacement was normalized to the foot width. Analysis of variance was used to calculate differences in CoP between test feet, and a comparison-wise alpha of P < 0.1 was considered as significant. Results In the TTA group, there was a significant difference between feet in A-P CoP location during early midstance (15% gait cycle). For the MX foot, the CoP was located at 21.3% ± 3.1% of the foot length, which was significantly greater than the LP and JS feet (15.6% ± 2% and 14.1% ± 3.4% foot length, respectively). During the period from terminal stance to toe-off (35%–60% gait cycle), the split keel design of JS and LP feet had greater M-L movement compared with that of the MX foot. In the TFA group, there were no differences in CoP trajectory between test feet. The maximum A-P excursion for all prosthetic feet was significantly greater than control subjects (80.4% ± 0.5% and 69.1% ± 2.7% foot length, respectively). The maximum M-L excursion for all prosthetic feet was significantly lesser than control subjects (14% ± 3.7% and 26.2% ± 3.7% foot width, respectively). Conclusions Differences in ankle design among different DR prosthetic feet can influence the asymmetry and functions of the locomotor system. In the TTA group, the MX foot resulted in a faster transition from initial contact to midstance, compared with the JS and LP feet. The split keel design of JS and LP feet closely mimicked the motion of the anatomical foot in late-stance period and provided controlled weight transfer from the prosthetic limb to the intact limb. In the TFA group, ankle design did not have a significant effect on CoP movement. All prosthetic feet had greater A-P CoP excursion and lower M-L excursion compared with the anatomical foot. Additional studies investigating the effects of DR foot/ankle designs on distinct functions of the locomotor system are needed to guide prosthetic foot prescription.

Using the loading response peak for defining gait cycle timing: A novel solution for the double-belt problem

Split-belt treadmills (SBTM) contain force plates under each belt that measure ground reaction force (GRF). Initial contact (IC) detection for each gait cycle obtained from the GRF is used for calculating temporal gait parameters (e.g., gait variability, step time, stride time). Occasionally, the participant steps with one leg on the contralateral belt (i.e., crossing) making the IC undetectable and the calculation of temporal gait parameters are compromised. We term this the double-belt problem (DBP). Objective: here we developed a complementary detection method using the loading response peak (LRP), anchor point for calculating gait parameters. Methods: we used GRF gait data from twenty adults (age 56.45 ± 4.81 y; 6 males) who walked on an SBTM. First, we used no-crossing gait periods free of the DBP to calculate stride time, step time, and stride time to stride time coefficient of variation and evaluated the true error and the normalized true error of the LRP detection method. Then, we used multiple comparisons between no-crossing data and crossing data. Results: we found that normalized errors (in comparison to the IC method) are ≤5.1%. Strong correlations were found between gait parameters computed based on the two detection methods (Intraclass correlation coefficient ≥0.97; p ≤ 0.001). Conclusion: detecting gait cycle timing based on the LRP detection method is reliable for estimating temporal gait parameters, demonstrating high correspondence with the gold standard IC detection method.

Rearfoot-forefoot profile defined by vertical ground reaction forces during gait is altered in patients with unilateral intermittent claudication

Intermittent Claudication due to Peripheral Arterial Disease (PAD-IC) induces ischemic pain in exercising muscles, and therefore impaired gait. In a pathological context, the analysis of the Vertical component of Ground Reaction Force (VGRF) is frequently used to describe gait pattern. This paper aims to define gait profiles according to the relative difference between peaks of VGRF; a Rearfoot and a Forefoot profile revealing a more loading or push-off strategy. We evaluated 70 participants (24 with unilateral disease (Unilat-IC), 22 with bilateral disease (Bilat-IC) and 24 Controls) during a walk test on an instrumented treadmill. Results indicate that Unilat-IC patients present a Rearfoot-profile in both legs during the pain-free gait period, likely to stabilize their gait. With the onset and increase of pain, the asymptomatic leg changes for a Forefoot-profile. This asymmetrical pattern suggests that a compensatory mechanism occurred to unload the symptomatic (painful) leg, possibly creating an imbalance. In Bilat-IC and Controls subjects, a Forefoot-profile is found, with a symmetrical pattern. However, there is a trend for lower propulsive capacity in case of Bilat-IC due to ischemic pain, but patients did not have the ability to compensate as in Unilat-IC. Therefore, Bilat-IC should not be considered as a “double” Unilat-IC. This study highlights the existence of gait profiles based on VGRF in PAD-IC patients. These profiles are dependent on the type of disease. Analysis of these gait profiles can 1) provide a simple way to identify gait alterations and 2) participate in improving physical rehabilitation strategies in PAD-IC patients.

A damped bipedal inverted pendulum for human–structure interaction analysis

The bipedal inverted pendulum with damping has been adopted to simulate human–structure interaction recently. However, the lack of analysis and verification has provided motivation for further investigation. Leg damping and energy compensation strategy are required for the bipedal inverted pendulum to regulate gait patterns on vibrating structures. In this paper, the Hunt–Crossley model is adopted to get zeros contact force at touch down, while energy compensation is achieved by adjusting the stiffness and rest length of the legs. The damped bipedal inverted pendulum can achieve stable periodic gait with a lower energy input and flatter attack angle so that more gaits are available, compared to the template, referred to as spring-load inverted pendulum. The measured and simulated vertical ground reaction force-time histories are in good agreement. In addition, the dynamic load factors are also within a reasonable range. Parametric analysis shows that the damped bipedal inverted pendulum can achieve stable gaits of 1.6 to 2.4 Hz with a reasonable first harmonic dynamic load factor, which covers the normal walking step frequency. The proposed model in this paper can be applied to human–structure interaction analysis.

Comfort optimization of a new type of foot mechanism for lower extremity exoskeleton

In order to reduce the impact caused by the contact between the foot and the ground when wearing the lower extremity exoskeleton under the condition of high load, this paper proposed an exoskeleton foot mechanism for improving the foot comfort, and optimized the key index of its influence on the comfort. Firstly, the physical model of foot mechanism was established based on the characteristics of foot stress in gait period, and then the mathematical model of vibration was abstracted. The correctness of the model was verified by the finite element analysis software ANSYS. Then, this paper analyzed the influence of vibration parameters on absolute transmissibility based on vibration mathematical model, and optimized vibration parameters with MATLAB genetic algorithm toolbox. Finally, this paper took white noise to simulate the road elevation as the vibration input, and used the visual simulation tool Simulink in MATLAB and the vibration equation to construct the acceleration simulation model, and then calculated the vibration weighted root mean square acceleration value of the foot. The results of this study show that this foot comfort mechanism can meet the comfort indexes of vibration absorption and plantar pressure, and this paper provides a relatively complete method for the design of exoskeleton foot mechanism, which has reference significance for the design of other exoskeleton foot and ankle joint rehabilitation mechanism.

Impulsive control and stability analysis of biped robot based on virtual constraint and adaptive optimization

This article is concerned with the problem of asymptotically stable periodic walking for a five‐link underactuated biped robot. Because of instantaneous change of two legs and complex dynamics during the walking process, the robot can be regarded as a nonlinear impulsive system. In order to make impulsive control of the robotic system, it is modeled as a rigid kinematic chain with Lagrange equations, which is strong coupling and hybrid. A novel state feedback controller is proposed to generate a closed‐loop periodic walking gait and track the desired trajectory of each joint. This controller is designed based on virtual constraint and hybrid zero dynamics. When the holonomic constraint of each actuator is zeroed by the controller, it can be regarded as an output. By using this method, the computation of the standard 5 degree of freedom (DOF) robot model can be reduced to 1 DOF. In addition, an adaptive optimization scheme is further investigated, which solves the problem of model parameter uncertainty. Finally, simulations are presented to illustrate the effectiveness of the proposed method.

Motor activation is modulated by visual experience during cyclic gait observation: A transcranial magnetic stimulation study.

Transcranial magnetic stimulation (TMS) has been widely utilized to noninvasively explore the motor system during the observation of human movement. However, few studies have characterized motor cortex activity during periodic gait observation. Thus, this study examined the effects of an observer’s visual experience and/or intention to imitate on corticospinal excitability during the observation of another’s gait. Twenty-six healthy volunteers were included in this study and allocated to two different groups. Participants in the visual experience group had formal experience with gait observation (physical therapist training), while those in the control group did not. Motor-evoked potentials induced by TMS in the tibialis anterior and soleus muscles were measured as surrogates of corticospinal excitability. Participants were seated and, while resting, they observed a demonstrator’s gait or observed it with the intention to subsequently reproduce it. Compared with the resting state, cyclic gait observation led to significant corticospinal facilitation in the tibialis anterior and soleus muscles. However, this pattern of corticospinal facilitation in the measured muscles was not coupled to the pattern of crural muscle activity during actual gait and was independent of the step cycle. This motor cortex facilitation effect during gait observation was enhanced by the observer’s visual experience in a manner that was not step cycle-dependent, while the observer’s intent to imitate did not affect corticospinal excitatory input to either muscle. In addition, visual experience did not modulate corticospinal excitability in gait-related crural muscles. Our findings indicate that motor cortex activity during gait observation is not in line with the timing of muscle activity during gait execution and is modulated by an individual’s gait observation experience. These results suggest that visual experience acquired from repetitive gait observation may facilitate the motor system’s control on bipedal walking, but may not promote the learning of muscle activity patterns.

Changes in gait parameters during rehabilitation after total knee arthroplasty

Knee arthroplasty (KAP) is a cost-effective and reliable treatment for knee osteoarthritis with the predicted increase in demand. Clinical assessment systems for the effectiveness of rehabilitation do not fully reflect the recovery process, but the evaluation of biomechanical gait parameters is most objective. Objective: to study of the changes in the biomechanical gait parameters after KAP for osteoarthritis. Subjects and methods . A total of 54 patients (31 women and 23 men; mean age, 63.86+1.69 years) were examined an average of 3.94+1.40 months after KAP at the Ivanovo Regional Hospital for War Veterans. A control group consisted of 49 people (28 women and 21 men; mean age, 34.6+8.8 years). Walking performance was studied three times: at admission, one and two weeks after the end of a cycle. Gait parameters were recorded using a Stedis feedback walking simulator (OOO “Neurosoft”, Ivanovo, Russia) in the Assessment package (Registration certificate No. РЗН 2018/7458 dated 08.07.2018). The results were processed using the standard biomedical statistical methods at a significance level of 5%. Results and discussion . In the early recovery period, the patients’ walking after KAP was characterized by decreased speed, by the symptoms of unloading, and by a slight lower extremity functional asymmetry. The revealed phenomena were suggestive of both positive changes in the recovery process as a whole and the normalization of the temporal pattern of a step cycle. It was established that walking speed and the double support period of gait were not restored 3 months after KAP. The main biomechanical changes were non-specific and were associated with reduced walking speed. It was suggested that physical therapy should be aimed mainly at increasing the stride length when controlling dynamic loads. The criterion for the effectiveness of rehabilitation was to reduce asymmetry when walking.

Effects of supraspinal feedback on human gait: rhythmic auditory distortion

BackgroundDifferent types of sound cues have been used to adapt the human gait rhythm. We investigated whether young healthy volunteers followed subliminal metronome rhythm changes during gait.MethodsTwenty-two healthy adults walked at constant speed on a treadmill following a metronome sound cue (period 566 msec). The metronome rhythm was then either increased or decreased, without informing the subjects, at 1 msec increments or decrements to reach, respectively, a low (596 msec) or a high frequency (536 msec) plateaus. After 30 steps at one of these isochronous conditions, the rhythm returned to the original period with decrements or increments of 1 msec. Motion data were recorded with an optical measurement system to determine footfall. The relative phase between sound cue (stimulus) and foot contact (response) were compared.ResultsGait was entrained to the rhythmic auditory stimulus and subjects subconsciously adapted the step time and length to maintain treadmill speed, while following the rhythm changes. In most cases there was a lead error: the foot contact occurred before the sound cue. The mean error or the absolute mean relative phase increased during the isochronous high (536 msec) or low frequencies (596 msec).ConclusionThese results showed that the gait period is strongly “entrained” with the first metronome rhythm while subjects still followed metronome changes with larger error. This suggests two processes: one slow-adapting, supraspinal oscillator with persistence that predicts the foot contact to occur ahead of the stimulus, and a second fast process linked to sensory inputs that adapts to the mismatch between peripheral sensory input (foot contact) and supraspinal sensory input (auditory rhythm).

Quantifying normal and parkinsonian gait features from home movies: Practical application of a deep learning-based 2D pose estimator.

Gait movies recorded in daily clinical practice are usually not filmed with specific devices, which prevents neurologists benefitting from leveraging gait analysis technologies. Here we propose a novel unsupervised approach to quantifying gait features and to extract cadence from normal and parkinsonian gait movies recorded with a home video camera by applying OpenPose, a deep learning-based 2D-pose estimator that can obtain joint coordinates from pictures or videos recorded with a monocular camera. Our proposed method consisted of two distinct phases: obtaining sequential gait features from movies by extracting body joint coordinates with OpenPose; and estimating cadence of periodic gait steps from the sequential gait features using the short-time pitch detection approach. The cadence estimation of gait in its coronal plane (frontally viewed gait) as is frequently filmed in the daily clinical setting was successfully conducted in normal gait movies using the short-time autocorrelation function (ST-ACF). In cases of parkinsonian gait with prominent freezing of gait and involuntary oscillations, using ACF-based statistical distance metrics, we quantified the periodicity of each gait sequence; this metric clearly corresponded with the subjects' baseline disease statuses. The proposed method allows us to analyze gait movies that have been underutilized to date in a completely data-driven manner, and might broaden the range of movies for which gait analyses can be conducted.

Study on effects of spinal joint for running quadruped robots

Most legged animals use their flexible body and supporting muscles to produce power for their locomotion, resulting in superior mobility and fast motions. In reality, an animal body consists of multiple bones and joints as well as legs having two or three segments with mass and inertia. In this paper, we study the bounding locomotion of a quadruped robot with a model closer to a real animal, i.e., a model that has one spinal joint, multiple two-segmented prismatic legs with masses and series elastic actuators, to obtain an insight into the robot’s dynamic behaviors. The models with passive mechanical properties are optimized with open-loop control to achieve the periodic bounding gait. The effects of spine flexibility in a segmented body are investigated on quadrupedal bounding gait by changing dynamic properties and hardware parameters. Comparisons of models reveal that body flexibility affects energy consumption and increases leg recirculation and stride length. The cost of transport of the articulated spine models is smaller than that of the rigid body one at low speed ($$ 0.45\sqrt{gl_0}$$). The stride length increases 25%. Furthermore, the study on location of spinal joint reveals that the asymmetric segmented body possesses bigger spine oscillation; up to 370% higher actuator force/torque in the rear leg but 36.1% smaller in the front leg; shorter stride period; and smaller cost of transport which helps the robot to run more efficiently. The study also shows that the asymmetric mass distribution of the body caused the torque/force increase at the rear leg, especially at hip joint, and the decrease at the front leg.

From standing balance to walking: A single control structure for a continuum of gaits

This article presents a control algorithm framework with which a bipedal robot can perform a variety of gaits by only modifying a small set of control parameters. The controller drives a number of variables, called non-emergent variables, to their desired trajectories resulting in a desired emergent walking gait. While the non-emergent variables remain the same independent of the gait, their desired trajectories are functions of a small set of control parameters that change as a function of the desired gait. This control algorithm has been tested on the humanoid robot COMAN, where different gaits including standing balance, stepping in place, periodic walking gaits with different velocities, as well as gait switching are demonstrated in experiments.

On the Existence and Uniqueness of Poincaré Maps for Systems With Impulse Effects

The Poincare map is widely used to study the qualitative behavior of dynamical systems. For instance, it can be used to describe the existence of periodic solutions. The Poincare map for dynamical systems with impulse effects (SIEs) was introduced in the last decade and mainly employed to study the existence of limit cycles (periodic gaits) for the locomotion of bipedal robots. We investigate sufficient conditions for the existence and uniqueness of Poincare maps for dynamical SIEs evolving on a differentiable manifold. We apply the results to show the existence and uniqueness of Poincare maps for systems with multiple domains.

Hierarchical and Safe Motion Control for Cooperative Locomotion of Robotic Guide Dogs and Humans: A Hybrid Systems Approach

This letter presents a hierarchical control strategy based on hybrid systems theory, nonlinear control, and safety-critical systems to enable cooperative locomotion of robotic guide dogs and visually impaired people. We address high-dimensional and complex hybrid dynamical models that represent collaborative locomotion. At the high level of the control scheme, local and nonlinear controllers, based on the virtual constraints approach, are designed to induce exponentially stable dynamic gaits. The local controller for the leash is assumed to be a nonlinear controller that keeps the human in a safe distance from the dog while following it. At the lower level, a real-time quadratic programming (QP) is solved for modifying the local controllers of the robot as well as the leash to avoid obstacles. In particular, the QP framework is set up based on control barrier functions (CBFs) to compute optimal control inputs that guarantee safety while being close to the local controllers. The stability of the complex periodic gaits is investigated through the Poincare return map. To demonstrate the power of the analytical foundation, the control algorithms are transferred into an extensive numerical simulation of a complex model that represents cooperative locomotion of a quadrupedal robot, referred to as Vision 60, and a human model. The complex model has 16 continuous-time domains with 60 state variables and 20 control inputs.

Subthalamic nucleus oscillations correlate with vulnerability to freezing of gait in patients with Parkinson's disease

Freezing of gait (FOG) is a disabling clinical phenomenon often found in patients with advanced Parkinson's disease (PD). FOG impairs motor function, causes falls and leads to loss of independence. Whereas dual tasking that distracts patients' attention precipitates FOG, auditory or visual cues ameliorate this phenomenon. The pathophysiology of FOG remains unclear. Previous studies suggest that the basal ganglia are involved in the generation of FOG. Investigation of the modulation of neuronal activities within basal ganglia structures during walking is warranted. To this end, we recorded local field potentials (LFP) from the subthalamic nucleus (STN) while PD patients performed single-task gait (ST) or walked while dual-tasking (DT). An index of FOG (iFOG) derived from trunk accelerometry was used as an objective measure to differentiate FOG-vulnerable gait from normal gait. Two spectral activities recorded from the STN region were associated with vulnerability to freezing. Greater LFP power in the low beta (15–21 Hz) and theta (5–8 Hz) bands were noted during periods of vulnerable gait in both ST and DT states. Whereas the elevation of low beta activities was distributed across STN, the increase in theta activity was focal and found in ventral STN and/or substantia nigra (SNr) in ST. The results demonstrate that low beta and theta band oscillations within the STN area occur during gait susceptible to freezing in PD. They also add to the evidence that narrow band ~18 Hz activity may be linked to FOG.

The effect of head movement on the bounding gait of a quadruped robot with an active spine

It is a common phenomenon in the movement of the quadruped mammals accompanied with head swings. Inspired by this, this article attempts to add head motion to the bounding gait of a quadruped robot. According to the theoretical analysis, there are two main functions of the head. First, the head can realize the active control of the center of mass position of the robot, which is of great significance to the stable motion of the robot. Second, the swing of the head plays a role in regulating the pitch angle of the torso and improves the coordination and stability of the motion. A simplified quadruped robot model with a head and spine joint is established and analyzed theoretically. The regularity of head movement in periodic bounding gait is summarized. Through simulation and experiment, we confirm the two roles of the head in the bounding gait of a quadruped robot.

Finite-time stabilization of periodic orbits for under-actuated biped walking with hybrid zero dynamics

This paper is concerned with the problem of finite-time stable walking for a 5-link under-actuated biped robot. Due to instantaneous change of 2 legs and complex dynamics during the walking process, the robot can be regarded as a nonlinear impulsive system. In order to make impulsive control of the robotic system, it is modeled as a rigid kinematic chain with Lagrange equations, which is strong coupling and hybrid. A novel finite-time feedback controller is proposed to realize finite-time stability of the nonlinear impulsive system. The controller is designed based on finite-time stabilizing control Lyapunov function (FTS-CLF) and hybrid zero dynamics (HZD). By establishing a finite-time stable periodic orbit, the controller can make the output of virtual constraints converge to zero rapidly. Restricted Poincare return map is then utilized to analyze the finite-time stability of nonlinear impulsive system. It ensures that the flow of the continuous subsystem can pass through the impact cross section. Additionally, a periodic walking gait planning is further investigated, and the existence of the gait is proved, which satisfy the joint trajectory tracking. Finally, the effectiveness of the mentioned method is illustrated by simulations.

Method based on UWB for user identification during gait periods

Everyone has a different way of walking, and for this reason, gait has been studied in the last few years as an important biometric information source. This study explores a novel approach, based on ultra-wideband (UWB) technology, for user identification via gait analysis. In the proposed method, the user is supposed to wear two or more devices embedding a UWB transceiver. During gait, the distances between the devices are estimated via UWB and then analysed by means of a machine learning classifier, which provides automatic identification. Experiments were carried out by 12 volunteers, who walked while wearing four UWB boards (placed on the head, wrist, ankle, and in a trouser pocket). The off-line evaluation considered a set of different possible configurations in terms of number and position of the wearable devices. Despite a relatively low sampling frequency of 10 Hz, the results are promising: average identification accuracy is as high as ∼96% with four devices, and above 90% with three devices (wrist, trouser pocket, and ankle). This novel approach may enhance the accuracy of inertial-based systems for continuous user identification.