Human Gait Analysis Research Articles

Non-Invasive Multi-Camera Gait Analysis System and its Application to Gender Classification

Objective: Most studies have conducted human gait analysis using expensive and invasive photogrammetric systems. The objective of this study was to demonstrate that non-invasive and cost-effective systems based on depth cameras may be able to retrieve relevant features of human gait patterns. We aimed to prove this by solving the problem of gait classification by gender. Methods: 81 participants (40 female and 41 male) walked at a self-selected speed across a 4.8-meter walkway. Gait data was recorded using multiple depth sensors. Analysis in time domain included joint excursions by gait phases, range of movement (ROM), central tendency and dispersion measures, spatial variables, and center of mass (COM) position. The spectral analysis included principal frequency, magnitude, and phase shift during walking. Only features with significant differences by gender were used to train a support vector machine (SVM) classifier. Results: A total of 108 features presented significant differences by gender (p<; 0.05). On this basis, the accuracy of the chosen model was 96.7%. Trunk rotation, trunk sway, knee abduction/adduction, and pelvic obliquity were the most differentiated between the groups. The COM position shown a significant difference by gender (p=0.0065) with 51.7% and 51.0% for men and women respectively. Women proved to have significantly shorter normalized step width than men (p=0.0472). Conclusion: The proposed method was able to retrieve most of human gait features correctly, including differences in gait pattern by gender. Significance: Depth cameras represent a cost-effective system that could be used for a deeper biomechanical human gait analysis.

Assessment of gait stability and preferred walking speed in virtual reality

Analysis of human gait as well as diagnosis of human locomotion organ should always be conducted with velocity of gait equal to Preferred Walking Speed (PWS). The literature review shows that the PWS value is not the same in real and virtual environment. The aim of this study was to determine PWS values in both environments and to specify values of parameters used in equations enabling PWS calculations on the basis of lower limb length. Research-related tests involved 40 subjects walking on the treadmill and wearing HMD goggles. The spatial scenery made participants feel like during a walk in the park. The tests included measurements of displacements of the COP, allowing for the calculation of the Lyapunov exponent and Floquet Multiplier. Both coefficients were used to identify stability at various gait velocities. The analysis revealed that the PWS in relation to gait on the treadmill with VR was lower than the PWS without VR. The final stage of research involved the determination of new values of coefficients of the formula enabling the identification of the velocity of comfort of gait in VR. Obtained results proved that PWS in real and virtual environment are different. The lower values were obtained for measurements in VR. On the basis of these results, value of the "a" coefficient, used in PWS calculations on the basis of lower limb length, was re-determined. The new value makes it possible to assess PWS for gait conducted on treadmill in virtual environment, what can be very important in gait evaluation.

Human Gait Recognition Using Hip, Knee and Ankle Joint Ratios

The estimation of joint angle ratios for healthy and afflicted subjects in characterizing the human gait has great significance in the development of limb prosthetics. The two dimensional analysis of human gait was performed and the ratio of hip to knee, knee to ankle, hip to ankle as well as the time taken for achieving a gait were determined. The percentage of affliction was computed based on the joint angle ratios and comparison was made with healthy gait. The joint ratios were fed as input to the driving system which comprises of six DC motors for the positioning of knee, hip and ankle during gait. Then different control strategies like P, PI and PID were tested. The t-test and ANOVA analysis were conducted between healthy, afflicted and PID controller to determine the significant difference between their joint angle ratios. The estimation of joint angle ratio improved the accuracy of the control system drive (desired position of knee, hip and ankle motors). The presence of oscillations in the output response was reduced for P and PI controllers. The implementation of PID controller eliminated the presence of peak overshoot and more settling time. Thus the joint angle ratio provides the best possible assistance to the disabled persons by appropriately compensating the affliction.

Input representations and classification strategies for automated human gait analysis

BackgroundQuantitative gait analysis produces a vast amount of data, which can be difficult to analyze. Automated gait classification based on machine learning techniques bear the potential to support clinicians in comprehending these complex data. Even though these techniques are already frequently used in the scientific community, there is no clear consensus on how the data need to be preprocessed and arranged to assure optimal classification accuracy outcomes. Research questionIs there an optimal data aggregation and preprocessing workflow to optimize classification accuracy outcomes? MethodsBased on our previous work on automated classification of ground reaction force (GRF) data, a sequential setup was followed: firstly, several aggregation methods – early fusion and late fusion – were compared, and secondly, based on the best aggregation method identified, the expressiveness of different combinations of signal representations was investigated. The employed dataset included data from 910 subjects, with four gait disorder classes and one healthy control group. The machine learning pipeline comprised principle component analysis (PCA), z-standardization and a support vector machine (SVM). ResultsThe late fusion aggregation, i.e., utilizing majority voting on the classifier's predictions, performed best. In addition, the use of derived signal representations (relative changes and signal differences) seems to be advantageous as well. SignificanceOur results indicate that great caution is needed when data preprocessing and aggregation methods are selected, as these can have an impact on classification accuracies. These results shall serve future studies as a guideline for the choice of data aggregation and preprocessing techniques to be employed.

Liquid State Machine to Generate the Movement Profiles for the Gait Cycle of a Six Degrees-of-Freedom Bipedal Robot in a Sagittal Plane

Abstract In this paper, an approach based on a liquid state machine (LSM) to compute the movement profiles to achieve a gait pattern subject to different variations in its trajectory is presented. At the same time, the position of the zero moment point (ZMP) to determine the stability of the six degrees-of-freedom (6DOF) bipedal robot in the sagittal plane during the gait cycle is calculated. The system is constructed as a supervised machine learning model. The time series of the oscillating foot trajectory obtained by direct kinematics with a multilayer perceptron neural network (MLP), to strengthen the kinematic model, is considered as input values for training. The target movement profiles are acquired of a human gait cycle analysis in three different scenarios: normal gait, climbing stairs, and descending stairs. In training, this model also gets the trajectories of the ZMP position during the gait cycle, as target time series. The LSM formed by spiking neurons, considered as third-generation neural networks, is compared in the accuracy of prediction, by the dynamic time warping (DTW) technique and correlation analysis, against the human gait analysis database. With this neuronal system, the joint positions to generate a trajectory of the oscillating foot and the ZMP position of the bipedal in the sagittal plane in different scenarios are obtained, proving the robustness of the LSM.

Learn to Adapt to Human Walking: A Model-Based Reinforcement Learning Approach for a Robotic Assistant Rollator

In this letter, we tackle the problem of adapting the motion of a robotic assistant rollator to patients with different mobility status. The goal is to achieve a coupled human–robot motion in a front-following setting as if the patient was pushing the rollator himself/herself. To this end, we propose a novel approach using model-based reinforcement learning (MBRL) for adapting the control policy of the robotic assistant. This approach encapsulates our previous work on human tracking and gait analysis from RGB-D and laser streams into a human-in-the-loop decision making strategy. We use long short-term memory (LSTM) networks for designing a human motion intention model and a coupling parameters forecast model, leveraging on the outcome of human gait analysis. An initial LSTM-based policy network was trained via imitation learning from human demonstrations in a motion capture setup. This policy is then fine-tuned with the MBRL framework using tracking data from real patients. A thorough evaluation analysis proves the efficiency of the MBRL approach as a user-adaptive controller.

Influence of musculotendon geometry variability in muscle forces and hip bone-on-bone forces during walking.

Inverse dynamics problems are usually solved in the analysis of human gait to obtain reaction forces and moments at the joints. However, these actions are not the actual forces and moments supported by the joint structure, because they do not consider the forces of the muscles acting across the joint. Therefore, to analyse bone-on bone forces it is necessary to estimate those muscle forces. Usually, this problem is addressed by means of optimization algorithms. One of the parameters required to solve this problem is the musculotendon geometry. These data are usually taken from cadavers or MRI data from several subjects, different from the analysed subject. Then, the model is scaled to the subject morphology. This procedure constitutes a source of error. The goals of this work were two. First, to perform a sensitivity analysis of the influence of muscle insertion locations on the muscle forces acting on the hip joint and on the hip joint bone-on-bone forces. Second, to compare the hip joint bone-on-bone forces during gait cycle obtained through muscle insertion locations taken from a musculoskeletal model template and a scaling procedure to those obtained from a subject-specific model using an MRI of the subject. The problem was solved using OpenSim. Results showed that anatomical variability should be analysed from two perspectives. One the one hand, throughout the gait cycle, in a global way. On the other hand, at a characteristic instant of the gait cycle. Variations of ±1 cm in the position of the attachment points of certain muscles caused variations of up to 14.21% in averaged deviation of the muscle forces and 58.96% in the peak force in the modified muscle and variations up to 57.23% in the averaged deviation of the muscle force and up to 117.23% in the peak force in the rest of muscles. Then, the influence of that variability on muscle activity patterns and hip bone-on-bone forces could be described more precisely. A biomechanical analysis of a subject-specific musculoskeletal model was carried out. Using MRI data, variations up to 5 cm in the location of the insertion points were introduced. These modifications showed significant differences between the baseline model and the customized model: within the range [-12%, 10%] for muscle forces and around 35% of body weight for hip bone-on-bone forces.

Gait analysis for early Parkinson’s disease detection based on deep learning

Abstract Better handling of neurological or neurodegenerative disorders such as Parkinson’s Disease (PD) is only possible with an early identification of relevant symptoms. Although the entire disease can’t be treated but the effects of the disease can be delayed with proper care and treatment. Due to this fact, early identification of symptoms for the PD plays a key role. Recent studies state that gait abnormalities are clearly evident while performing dual cognitive tasks by people suffering with PD. Researches also proved that the early identification of the abnormal gaits leads to the identification of PD in advance. Novel technologies provide many options for the identification and analysis of human gait. These technologies can be broadly classified as wearable and non-wearable technologies. As PD is more prominent in elderly people, wearable sensors may hinder the natural persons movement and is considered out of scope of this paper. Non-wearable technologies especially Image Processing (IP) approaches captures data of the person’s gait through optic sensors Existing IP approaches which perform gait analysis is restricted with the parameters such as angle of view, background and occlusions due to objects or due to own body movements. Till date there exists no researcher in terms of analyzing gait through 3D pose estimation. As deep leaning has proven efficient in 2D pose estimation, we propose an 3D pose estimation along with proper dataset. This paper outlines the advantages and disadvantages of the state-of-the-art methods in application of gait analysis for early PD identification. Furthermore, the importance of extracting the gait parameters from 3D pose estimation using deep learning is outlined.

Deep Learning for Monitoring of Human Gait: A Review

The essential human gait parameters are briefly reviewed, followed by a detailed review of the state of the art in deep learning for the human gait analysis. The modalities for capturing the gait data are grouped according to the sensing technology: video sequences, wearable sensors, and floor sensors, as well as the publicly available datasets. The established artificial neural network architectures for deep learning are reviewed for each group, and their performance are compared with particular emphasis on the spatiotemporal character of gait data and the motivation for multi-sensor, multi-modality fusion. It is shown that by most of the essential metrics, deep learning convolutional neural networks typically outperform shallow learning models. In the light of the discussed character of gait data, this is attributed to the possibility to extract the gait features automatically in deep learning as opposed to the shallow learning from the handcrafted gait features.

Introduction of Triboelectric Positive Bioplastic for Powering Portable Electronics and Self-Powered Gait Sensor

Triboelectric nanogenerators (TENGs) are in the center of alternative energy scavenger technologies because of numerous possible applications in remote energy harvesting and self-powered sensing. Although there exists a large number of triboelectric negative polymers, the choice of triboelectric positive polymers is indeed limited. Here we report for the first time the use of thermoplastic starch (TPS) as a triboelectric positive material for designing TENGs. Our bioplastic-based highly durable TENG (b-TENG) can generate open-circuit peak-to-peak output voltage of ∼560 V with output current density of ∼120 mA m–2 and instantaneous output power density of 17 W m–2. We demonstrate b-TENG as a portable power source by powering more than 100 commercial blue LEDs, LED strips, and seven segment LCD screens. Additionally, the b-TENG can be used as a self-powered pedometer for step counter, a speedometer for human walking and running, and a human gait analysis sensor to assess physical activity breeding its futuristic biomedical applications for healthcare purposes. The introduction of eco-friendly bioplastic TPS as a triboelectric positive component has great potential for biomedical applications because of the abundance, biodegradability, low cost, and ease in fabrication process.

Using Smartphone Inertial Measurement Unit for Analysis of Human Gait

The study was aimed at searching the characteristic features of a human gait during a conventionally used neurologic walking test with the help of sensors (3D accelerometer and gyroscope) in a smartphone mounted on the person's head. This allowed reducing the amount of analyzed data and saving time for analysis in comparison with motion video capture methods. It has been found that merely one inertial unit is good enough to detect the gait left-right asymmetry in healthy subjects. Several parameters were derived, one of them (acceleration shock) proved to be the most informative one. The gyroscope signal allowed detecting the characteristics of the gait in Parkinson's disease patients seen as the excess of the power spectrum density in all three axes. Therefore, this study was aimed to extract amplitude and spectral parameters from acceleration and rotation rate signals of smartphone-based IMU attached to the head during the long version of the TUG test in a group of neurologically healthy young subjects.

2D Gait Analysis for Biomedical Application

The paper presents preparation of 2D lab setup for human gait analysis. In the system 2D human gait lab is setup by using cameras in front and sagittal side. Parallax error has minimize by considering gait factors during calculation. 3D human gait analysis affords reliable result data, but presently in narrow use because of high instrumental arrangement and their maintenance costs. 2D Indian scientific community can advantage from traditional style approaches which are easy to implement, low-cost, reliable in all variety of set-ups with relative ease.

A review of simulation methods for human movement dynamics with emphasis on gait

Human gait analysis is a complex problem in biomechanics because of highly nonlinear human motion equations, muscle dynamics, and foot-ground contact. Despite a large number of studies in human gait analysis, predictive human gait simulation is still challenging researchers to increase the accuracy and computational efficiency for evaluative studies (e.g., model-based assistive device controllers, surgical intervention planning, athletic training, and prosthesis and orthosis design). To assist researchers in this area, this review article classifies recent predictive simulation methods for human gait analysis according to three categories: (1) the human models used (i.e., skeletal, musculoskeletal and neuromusculoskeletal models), (2) problem formulation, and (3) simulation solvers. Human dynamic models are classified based on whether muscle activation and/or contraction dynamics or joint torques (instead of muscle dynamics) are employed in the analysis. Different formulations use integration and/or differentiation or implicit-declaration of the dynamic equations. A variety of simulation solvers (i.e., semi- and fully-predictive simulation methods) are studied. Finally, the pros and cons of the different formulations and simulation solvers are discussed.

Explaining the unique nature of individual gait patterns with deep learning

Machine learning (ML) techniques such as (deep) artificial neural networks (DNN) are solving very successfully a plethora of tasks and provide new predictive models for complex physical, chemical, biological and social systems. However, in most cases this comes with the disadvantage of acting as a black box, rarely providing information about what made them arrive at a particular prediction. This black box aspect of ML techniques can be problematic especially in medical diagnoses, so far hampering a clinical acceptance. The present paper studies the uniqueness of individual gait patterns in clinical biomechanics using DNNs. By attributing portions of the model predictions back to the input variables (ground reaction forces and full-body joint angles), the Layer-Wise Relevance Propagation (LRP) technique reliably demonstrates which variables at what time windows of the gait cycle are most relevant for the characterisation of gait patterns from a certain individual. By measuring the time-resolved contribution of each input variable to the prediction of ML techniques such as DNNs, our method describes the first general framework that enables to understand and interpret non-linear ML methods in (biomechanical) gait analysis and thereby supplies a powerful tool for analysis, diagnosis and treatment of human gait.

Abnormal Gait Recognition Algorithm Based on LSTM-CNN Fusion Network

This paper presents a novel approach to human gait analysis with a sensor-based technique involving a wearable inertial measurement unit (IMU). The proposed system emphasizes the detection of certain abnormal gait patterns, including hemiplegic, tiptoe, and cross-threshold gait. First, we use the dynamic step conjugate gradient algorithm to calculate the attitude of the gait data, and we then use the gait feature information location algorithm to segment the attitude data. The segmented attitude data are used as input in the classification model. In this paper, we propose an algorithm based on a long short-term memory network and convolutional neural network (LCWSnet) for diagnosis and classification of abnormal gait patterns using the leg Euler angle information, and parameters related to features can be adjusted adaptively according to the feedback of objectives and optimization functions. We optimize the convergence layer of the LSTM-CNN model and improve the classification accuracy of abnormal gait. The experimental results demonstrate that the proposed LCWSnet-based technique is able to detect gait abnormality in the data. The proposed personalized gait classification approach is accurate and reliable and can be implemented for the abnormal gait.

Method of synchronization and data processing from differents inertial sensors kits sources for the human gait analysis

The article talks about results of data synchronization measurements sourced from two recording gait systems for human gait analyses. Two systems are Xsens sensor kits: MT Awinda, Xbus Kit. The article cover file format used to save data and synchronization method for sensor measurement from above mentioned kits. On the basis of the studies carried out, sensor measurement from different places on human body are unify to a common frame of reference. The discussed method provides also progressive data processing for angles range from -180° to 180° conversion to the absolute angle value from initial sensor settings.

A survey of wearable sensor networks in health and entertainment

The world of wearable sensor networks is constantly improving as the potential benefits these systems offer become more apparent Today there are several wearable sensor networks being developed for health and wellness purposes as well as entertainment purposes This paper will take a closer look at the different available sensors that are being used to collect data as well as the systems that use them Sensors for fall protection for the elderly long term health monitoring human gait analysis interactive media and gaming and animation are studied with respect to the state of the art in wearable sensor technologies This research hopes to further advance the technology for the wearable sensor networks and wearable health monitoring systems in these fields

System for automatic gait analysis based on a single RGB-D camera

Human gait analysis provides valuable information regarding the way of walking of a given subject. Low-cost RGB-D cameras, such as the Microsoft Kinect, are able to estimate the 3-D position of several body joints without requiring the use of markers. This 3-D information can be used to perform objective gait analysis in an affordable, portable, and non-intrusive way. In this contribution, we present a system for fully automatic gait analysis using a single RGB-D camera, namely the second version of the Kinect. Our system does not require any manual intervention (except for starting/stopping the data acquisition), since it firstly recognizes whether the subject is walking or not, and identifies the different gait cycles only when walking is detected. For each gait cycle, it then computes several gait parameters, which can provide useful information in various contexts, such as sports, healthcare, and biometric identification. The activity recognition is performed by a predictive model that distinguishes between three activities (walking, standing and marching), and between two postures of the subject (facing the sensor, and facing away from it). The model was built using a multilayer perceptron algorithm and several measures extracted from 3-D joint data, achieving an overall accuracy and F1 score of 98%. For gait cycle detection, we implemented an algorithm that estimates the instants corresponding to left and right heel strikes, relying on the distance between ankles, and the velocity of left and right ankles. The algorithm achieved errors for heel strike instant and stride duration estimation of 15 ± 25 ms and 1 ± 29 ms (walking towards the sensor), and 12 ± 23 ms and 2 ± 24 ms (walking away from the sensor). Our gait cycle detection solution can be used with any other RGB-D camera that provides the 3-D position of the main body joints.

An Omnidirectional Biomechanical Energy Harvesting (OBEH) Sidewalk Block for a Self-Generative Power Grid in a Smart City

Energy harvesting, which converts ambient, otherwise wasted, energy sources into usable electricity, is expected to contribute to the formulation of a self-generating power grid. This type of grid can enable sustainable operation of wireless sensor networks as the “Smart City” vision becomes reality. Human walking is a plentiful mechanical energy source wasted during daily activities. This study aims to develop an omnidirectional biomechanical energy harvesting (OBEH) sidewalk block that is able to generate electricity from human walking. Here, a systematic design framework for the OBEH sidewalk block is presented; it consists of three important ingredients, specifically: (1) extraction of a footstep loading profile from human gait analysis; (2) electroelastically coupled finite element modeling to estimate the transient output responses under the footstep loading profile; and (3) reliability-based design optimization of the OBEH sidewalk block. This study considers two kinds of the inherent randomness, including (1) variability in the material properties and geometry; and (2) uncertainty in the position and direction of the footsteps. It can be concluded from the results that the optimum design of the proposed OBEH sidewalk block enables useful power generation while satisfying the target reliability of fatigue failure in the presence of the inherent randomness.

Extraction and Classification of Human Body Parameters for Gait Analysis

Human gait analysis is considered a new biometric tool for the ability to obtain metrics from the body at a distance. Biometric identifiers have properties that can technologically measure and analyze the characteristics of the human body and can be used as a form of identification and access control for security applications. Recognition through proper interpretation of gait parameters has become a relevant pattern classification problem. This work aims to develop an image processing system, with the use of the Microsoft Kinect sensor, which is capable to extract movement patterns for gait analysis and to present a comparative study of different pattern recognition methods for human identification. The image processing system, developed in C#, allowed the acquisition of three-dimensional data from several volunteers and made it possible to identify the human skeleton and automatically extract the kinetic and kinematic parameters of the body. For data analysis, different classification methods were compared. Among them, the algorithms that presented better performance were probabilistic neural networks, deep neural networks and k-nearest neighbors, with nearly 99% correct recognition rate. The obtained results demonstrate the efficiency of gait analysis as a biometric method. They also show the viability of gait parameter extraction using the Kinect sensor and the good performance of pattern recognition methods applied to the acquired gait kinetic and kinematic parameters.