Gait Biometrics Research Articles

Deep supervised hashing for gait retrieval.

Background: Gait recognition is perceived as the most promising biometric approach for future decades especially because of its efficient applicability in surveillance systems. Due to recent growth in the use of gait biometrics across surveillance systems, the ability to rapidly search for the required data has become an emerging need. Therefore, we addressed the gait retrieval problem, which retrieves people with gaits similar to a query subject from a large-scale dataset. Methods: This paper presents the deep gait retrieval hashing (DGRH) model to address the gait retrieval problem for large-scale datasets. Our proposed method is based on a supervised hashing method with a deep convolutional network. We use the ability of the convolutional neural network (CNN) to capture the semantic gait features for feature representation and learn the compact hash codes with the compatible hash function. Therefore, our DGRH model combines gait feature learning with binary hash codes. In addition, the learning loss is designed with a classification loss function that learns to preserve similarity and a quantization loss function that controls the quality of the hash codes Results: The proposed method was evaluated against the CASIA-B, OUISIR-LP, and OUISIR-MVLP benchmark datasets and received the promising result for gait retrieval tasks. Conclusions: The end-to-end deep supervised hashing model is able to learn discriminative gait features and is efficient in terms of the storage memory and speed for gait retrieval.

Reliability and generalization of gait biometrics using 3D inertial sensor data and 3D optical system trajectories

Particularities in the individuals’ style of walking have been explored for at least three decades as a biometric trait, empowering the automatic gait recognition field. Whereas gait recognition works usually focus on improving end-to-end performance measures, this work aims at understanding which individuals’ traces are more relevant to improve subjects’ separability. For such, a manifold projection technique and a multi-sensor gait dataset were adopted to investigate the impact of each data source characteristics on this separability. Assessments have shown it is hard to distinguish individuals based only on their walking patterns in a subject-based identification scenario. In this setup, the subjects’ separability is more related to their physical characteristics than their movements related to gait cycles and biomechanical events. However, this study’s results also points to the feasibility of learning identity characteristics from individuals’ walking patterns learned from similarities and differences between subjects in a verification setup. The explorations concluded that periodic components occurring in frequencies between 6 and 10 Hz are more significant for learning these patterns than events and other biomechanical movements related to the gait cycle, as usually explored in the literature.

Effect of biologically-motivated energy constraints on liquid state machine dynamics and classification performance

Abstract Equipping edge devices with intelligent behavior opens up new possibilities for automating the decision making in extreme size, weight, and power-constrained application domains. To this end, several recent lines of research are aimed at the design of artificial intelligence hardware accelerators that have significantly reduced footprint and power demands compared to conventional CPU/GPU systems. However, despite some key advancements, the majority of work in this area assumes that there is an unlimited supply of energy available for computation, which is not realistic in the case of battery-powered and energy harvesting devices. In this paper, we address this gap by exploring the computational effects of energy constraints on a popular class of brain-inspired spiking neural networks–liquid state machines (LSMs). Energy constraints were applied by limiting the spiking activity in subsets of LSM neurons. We tested our designs on two biosignal processing tasks: epileptic seizure detection and biometric gait identification. For both tasks, we show that energy constraints can significantly improve classification accuracy. This demonstrates that in the design of neuromorphic systems, reducing energy and increasing performance are not always competing goals.

Biometric-Based Security System for Smart Riding Clubs

Horses, workers or riders need safety in a farm or a riding club. On account of the great value of the horse, the breeder needs to protect it from theft and disease. In this context, it is important to detect and recognize the identity of each worker or rider and horse for security reasons. In fact, this paper proposes a Smart Riding Club Biometric System (SRCBS) consisting in automatically detecting and recognizing horses as well as humans. The proposed system is based on the facial biometrics for a horse as well as the human gait biometrics due to their simplicity and intuitiveness in an uncontrolled environment. This work suggests a Siamese network based on DenseNet features for human gait recognition, a Sparse Neural Network (SNN) based on sparse features for horse face detection and a horse face recognition method based on Gabor features, LDA and SVM. Because of the unavailability of horse databases, this paper presents a new benchmark for horse detection and recognition in order to evaluate our proposed system. The proposed systems achieved an average accuracy of gait recognition equal to 95% in the 0° view, 100% in the 90° view and 98.90% in the 180° view on Casia-B dataset, an average precision equal to 90% for horse face detection and a recognition rate equal to 99.89% for horse face identification.

Requirements for the Procedure for Assessing the Operator’s Person When Processing Accelerometer Data

The article considers the possibility of biometric authentication based on gait parameters, which are obtained after intelligent processing of the accelerometer data of a wearable device. The article discusses the main trends and trends in the field of biometric authentication, as well as authentication by gait parameters. The developed neural network algorithm and informative parameters are described in the authentication procedure based on the data of a single sensor of a portable device. The practical verification of the proposed approach is carried out on 32 subjects of different physiology. The results of the study show the possibility of distinguishing their own movements in 100% of cases, and the distinction of the subjects is more than 90%. Also, the final part of the article provides the requirements for the authentication procedure when processing accelerometric data of gait biometrics, the level of trust of the developed algorithm is determined.

Deep supervised hashing for gait retrieval

Background: Gait recognition is perceived as the most promising biometric approach for future decades especially because of its efficient applicability in surveillance systems. Due to recent growth in the use of gait biometrics across surveillance systems, the ability to rapidly search for the required data has become an emerging need. Therefore, we addressed the gait retrieval problem, which retrieves people with gaits similar to a query subject from a large-scale dataset. Methods: This paper presents the deep gait retrieval hashing (DGRH) model to address the gait retrieval problem for large-scale datasets. Our proposed method is based on a supervised hashing method with a deep convolutional network. We use the ability of the convolutional neural network (CNN) to capture the semantic gait features for feature representation and learn the compact hash codes with the compatible hash function. Therefore, our DGRH model combines gait feature learning with binary hash codes. In addition, the learning loss is designed with a classification loss function that learns to preserve similarity and a quantization loss function that controls the quality of the hash codes Results: The proposed method was evaluated against the CASIA-B, OUISIR-LP, and OUISIR-MVLP benchmark datasets and received the promising result for gait retrieval tasks. Conclusions: The end-to-end deep supervised hashing model is able to learn discriminative gait features and is efficient in terms of the storage memory and speed for gait retrieval.

3D Stacked Near‐Field Electrospun Nanoporous PVDF‐TrFE Nanofibers as Self‐Powered Smart Sensing in Gait Big Data Analytics

AbstractThis study describes a new stacked structure with porous nanofiber‐based piezoelectric nanogenerator by near‐field electrospinning. The 3D‐stacked porous nanofibers structure enhances the stress concentration effect such that the PVDF‐TrFE nanofibers can promote higher performance in electrical output. By mimicking the additive manufacturing in near‐field electrospinning (NFES) setup, structurally simple and piezoelectrically effective fabrication is capable of converting stacked porous PVDF‐TrFE nanofiber into a high‐performance sensor. The electrical voltage output performance is enhanced more than 2.7 times compared with primitive PVDF‐TrFE nanofiber. Furthermore, a self‐powered foot pressure recognition statistical system and an individual gait biometrics system are developed to provide gait recognition and a new biometrics technology. A personal sequence gait piezoelectric signal recognition rate of 86% is achieved by deep learning BiLSTM model. Furthermore, besides expanding the application area of self‐powered system to smart wearable device monitoring, this work also stimulates the evolution of big data analytics in the intelligent medical industry.

A Fuzzy Authentication System Based on Neural Network Learning and Extreme Value Statistics

Internet-connected smart devices in, on, and around us, (e.g., embedded devices, wearable devices, and smart sensors) can collect human biometric features and facilitate identity authentication. Existing approaches are mainly based on pattern recognition and machine learning algorithms, which may not be capable of processing uncertain user information. Thus, focusing on the uncertainty of users' identity, this article proposes a fuzzy authentication system based on neural network and extreme value analysis. Specifically, we utilize biometric gait information of human body recognition. Our proposed authentication system is designed to implicitly authenticate users based on their gait, and can detect uncertain users and reject the authentication of unknown users. The performance is evaluated using an open dataset of 153 volunteers, where we manage to achieve a recognition accuracy rate of 98.4% and an error rate of unauthorized users at 6%.

Animação gráfica da marcha humana a partir de dados do Kinect

The analysis of human gait from biometric data has applications in areas such as safety, bio-inspired robotics and health. Low-cost motion sensors such as Kinect have enabled the acquisition of biometric gait data in terrestrial environments. However, these devices have limitations that can impact the quality of the acquired data. In this scenario, different signal processing techniques can be applied to reduce noise. Visualization of this data, whether original or processed, is often performed in the form of graphs, having limited utility for professionals not experienced in signal analysis. In this sense, the visualization of gait data in a three-dimensional model can contribute to better decision-making by professionals, especially in healthcare. This work aims to perform the graphic animation of the human gait in a three-dimensional model, from data collected by the Kinect 2.0 sensor. To reduce the noise of the acquired data, a pre-processing was performed with the application of moving average and Butterworth filters. Videos of the animations in isometric and side views were created and evaluated in a field research on the artificiality/naturalness of animation, using the mean opinion score (MOS) technique. A total of 22 participants answered the questionnaire. The one-way ANOVA statistical test showed that videos in isometric and side view processed with moving average filter (window=15 and repetitions=3) that obtained higher MOS metric values ​​were evaluated as significantly more natural than other processed videos or not.

Gait biometrics: investigating the use of the lower inner regions for people identification from landmark frames

The recent technological advances in surveillance, forensic and biometric systems to deter or even reduce the increasing number of crimes and prevent them is still questionable. The use of gait biometrics has attracted unprecedented interest due to its capability to work with low-resolution footage recorded from a distance. In contrast to mainstream research on gait biometrics which uses holistic silhouette features, the authors investigate the use of the bottom dynamic section within the human body to derive the most discriminative features for gait recognition. A new descriptor based on 7 Hu's moments is proposed describing the inner lower limb regions between the limbs being extracted only from landmark frames within one gait cycle. In order to assess the discriminatory potency of gait features from the lower regions for people identification, a number of experiments are conducted on the CASIA-B gait database to investigate the recognition rates using the KNN classifier and deep learning. The comparative analysis is performed against well-established research studies which were tested on the CASIA-B data set. The obtained results confirm the consistency of features extracted from the lower regions for gait recognition even under the impact of various factors.

On Learning Disentangled Representations for Gait Recognition.

Gait, the walking pattern of individuals, is one of the important biometrics modalities. Most of the existing gait recognition methods take silhouettes or articulated body models as gait features. These methods suffer from degraded recognition performance when handling confounding variables, such as clothing, carrying and viewing angle. To remedy this issue, we propose a novel AutoEncoder framework, GaitNet, to explicitly disentangle appearance, canonical and pose features from RGB imagery. The LSTM integrates pose features over time as a dynamic gait feature while canonical features are averaged as a static gait feature. Both of them are utilized as classification features. In addition, we collect a Frontal-View Gait (FVG) dataset to focus on gait recognition from frontal-view walking, which is a challenging problem since it contains minimal gait cues compared to other views. FVG also includes other important variations, e.g., walking speed, carrying, and clothing. With extensive experiments on CASIA-B, USF, and FVG datasets, our method demonstrates superior performance to the SOTA quantitatively, the ability of feature disentanglement qualitatively, and promising computational efficiency. We further compare our GaitNet with state-of-the-art face recognition to demonstrate the advantages of gait biometrics identification under certain scenarios, e.g., long-distance/lower resolutions, cross viewing angles. Source code is available at http://cvlab.cse.msu.edu/project-gaitnet.html.

Gait biometrics in children with cerebral palsy before and after robotic mechanotherapy

Background. The improvement in existing methods and the development of new principles for treating children with cerebral palsy necessitates a quantitative assessment of the parameters of motor activity. However, because of the explicit and complex abnormalities in motor skills in patients with severe forms of cerebral palsy, an evaluation of their locomotor function dynamics using instrumental diagnostic methods remains a serious problem.
 Aim. This work aimed to study the walking function in patients with cerebral palsy before and after motor rehabilitation using a biomechanical method with biometric sensors.
 Materials and methods. We examined 14 patients with cerebral palsy aged 8 to 13 years with III level of restriction of motor activity according to the gross motor function classification system (GMFCS). All patients underwent rehabilitation in the Lokomat robotic simulator for three weeks. The course consisted of 15 sessions of 45 min each. The temporal and dynamic parameters of walking were studied in 14 patients with cerebral palsy before and after a course of locomotor training. The biometry of the step cycle was studied using the STEDIS hardware-software complex, including the Neurosens set of wireless biometric sensors. The temporal characteristics of the step cycle and the force interaction of the lower extremities with the supporting surface during walking were recorded. For comparison, we conducted a biomechanical examination of 18 healthy children of the same age who did not have signs of orthopedic disorders.
 Results. Although after a course of mechanotherapy, the indices of the support phases in biometry in children with cerebral palsy did not reach the level of healthy individuals, a physiological tendency to roll foot was observed in the phase of pushing and accelerating the foot. Active braking of the lower limb increased. The studied time parameters showed a relative improvement in the step structure because of the emerging tendency to normalize the ratio of the periods of the double support of the contralateral lower extremities.
 Conclusion. Robotic mechanotherapy helps to change the biomechanical pattern of walking of a child with a severe degree of cerebral palsy. An instrumental analysis of walking using wireless biometric sensors allows you to evaluate the results and effectiveness of rehabilitation measures in patients with severe motor impairment objectively.

IMPORTANT FACTORS FOR EVALUATION OF BASED ANALYSIS SYSTEMS BASED ON BIOMETRIC RECOGNITION

Gait recognition to be the recognition of some salient property, e.g., identity, style of walk, or pathology, based on the coordinated, cyclic motions that result in human locomotion. In the case of biometric gait recognition, the salient property is identity. We make the distinction between gait recognition and what we call quasi gait recognition in which a salient property is recognized based on features acquired while a subject is walking, but the features are not inherently part of the gait. For example, skeletal dimensions may be measured during gait and used to recognize an individual.

Deep Learning-Based Gait Recognition Using Smartphones in the Wild

Compared to other biometrics, gait is difficult to conceal and has the advantage of being unobtrusive. Inertial sensors, such as accelerometers and gyroscopes, are often used to capture gait dynamics. These inertial sensors are commonly integrated into smartphones and are widely used by the average person, which makes gait data convenient and inexpensive to collect. In this paper, we study gait recognition using smartphones in the wild. In contrast to traditional methods, which often require a person to walk along a specified road and/or at a normal walking speed, the proposed method collects inertial gait data under unconstrained conditions without knowing when, where, and how the user walks. To obtain good person identification and authentication performance, deep-learning techniques are presented to learn and model the gait biometrics based on walking data. Specifically, a hybrid deep neural network is proposed for robust gait feature representation, where features in the space and time domains are successively abstracted by a convolutional neural network and a recurrent neural network. In the experiments, two datasets collected by smartphones for a total of 118 subjects are used for evaluations. The experiments show that the proposed method achieves higher than 93.5\% and 93.7\% accuracies in person identification and authentication, respectively.

A Survey of Behavioral Biometric Gait Recognition: Current Success and Future Perspectives

In today digital society, vulnerability to person authentication is a serious issue in real time scenarios like (airport, hospital, metro stations, etc.). This issue has increased the growth of video surveillance security systems. In recent decades behavioral biometric trait gait has emerged as a potential surveillance monitoring system because of its inconspicuous and unperceivable nature. Even more human gait has a benefit that it can be tracked at a distance and under low resolution videos. Finally, it is difficult to impersonate gait features. In this article, we comprehensively investigate the past and current research development in vision-based (VB) gait recognition. We give a brief description of feature selection and classification techniques used in gait recognition. The article extensively investigates feature representation techniques, classified into model-based and model-free. The article also provides a detail description of databases that are available for research purposes classified into two categories: VB and sensor-based. We extensively examine factors that affect gait recognition, and current research was done to cope with these factors. Moreover, this article proposes future perspectives after investigating state-of-art literature that can be more helpful to experts and new comers in gait recognition. In last, we also give a brief description of the proposed workflow.

Use of a pressure-sensing walkway system for biometric assessment of gait characteristics in goats

The purpose of this study was to quantitatively assess gait characteristics and weight-bearing forces during ambulation in goats free of lameness using a pressure-sensing walkway as a biometric tool for stride, gait, and force analysis. Forty-six non-lame adult goats ranging in age from 5 to 6 years, mixed-breeds, and with a mean body weight of 52 ± 7.1 kgs were used. Goats were trained to walk over a pressure-sensing walkway. Data for analysis was collected on 2 different days, 3 days apart. On each day, 2 to 5 walking passes, in the same direction, were captured for each goat. Data from 2 valid passes meeting the criteria for consistent walking gait on each day were averaged then used for analysis. Analysis was performed, including the day-effect, for stride, gait, and force characteristics. Of the 46 goats enrolled in the study, complete data sets were achieved in 33 (72%) goats. Gait biometrics were similar among the assessment days; therefore, all data was pooled for the purpose of characterizing data for individual limb and biometric parameter comparisons at the individual goat level. Statistical analysis revealed that no difference within the paired limbs, and that there were significant differences between the front limbs and hind limbs. Maximum force and maximum peak pressure were significantly greater for the front limbs as compared with the hind limbs (p < 0.001). Based on the results, gait and force characteristics can be consistently measured in goats using a pressure-sensing walkway during a consistent walking gait. Goats apply greater force to the forelimbs during the weight-bearing phase of stride as compared with the hind limbs. The use of objective assessment tools is expected to improve the ability of researchers and clinicians to monitor changes in weight bearing and gait and will contribute to improved animal welfare.

Calibrated and synchronized multi-view video and motion capture dataset for evaluation of gait recognition

We introduce synchronized and calibrated multi-view video and motion capture dataset for motion analysis and gait identification. The 3D gait dataset consists of 166 data sequences with 32 people. In 128 data sequences, each of 32 individuals was dressed in his/her clothes, in 24 data sequences, 6 of 32 performers changed clothes, and in 14 data sequences, 7 of the performers had a backpack on his/her back. In a single recording session, every performer walked from right to left, then from left to right, and afterwards on the diagonal from upper-right to bottom-left and from bottom-left to upper-right corner of a rectangular scene. We demonstrate that a baseline algorithm achieves promising results in a challenging scenario, in which gallery/training data were collected in walks perpendicular/facing to the cameras, whereas the probe/testing data were collected in diagonal walks. We compare performances of biometric gait recognition that were achieved on marker-less and marker-based 3D data. We present recognition performances, which were achieved by a convolutional neural network and classic classifiers operating on gait signatures obtained by multilinear principal component analysis. The availability of synchronized multi-view image sequences with 3D locations of body markers creates a number of possibilities for extraction of discriminative gait signatures. The gait data are available at http://bytom.pja.edu.pl/projekty/hm-gpjatk/.

A comprehensive study on gait biometrics using a joint CNN-based method

This paper gives a comprehensive study on gait biometrics via a joint CNN-based method. Gait is a kind of behavioral biometric feature with unique advantages, e.g., long-distance, cross-view and non-cooperative perception and analysis. In this paper, the definition of gait analysis includes gait recognition and gait-based soft biometrics such as gender and age prediction. We propose to investigate these two problems in a joint CNN-based framework which has been seldom reported in the recent literature. The proposed method is efficient in terms of training time, testing time and storage. We achieve the state-of-the-art performance on several gait recognition and soft biometrics benchmarks. Also, we discuss which part of the human body is important and informative for a specific task by network visualization.

A Smart Security System Using Multimodal Features from Videos

Multi-modal biometric authentication system uses more than one biometric feature and the use of multi-modal biometrics improves security by making the system invulnerable to spoofing attacks. The proposed system uses face and gait biometrics for authentication and identification. The authentication is done in an unobtrusive manner without the knowledge and co-ordination of the user with the help of surveillance cameras. The videos are captured from two surveillance cameras, placed at fronto-parallel and fronto-normal views are given as input to the system. The gait system uses the video from fronto-parallel view and uses a model free approach to extract a spatio temporal motion summary of the gait cycle. The gait features has been compared by calculating the Euclidean distance between them. The face system uses the video from fronto normal view and uses an appearance based approach to extract features from the face of the user. The face features has been compared by calculating Chi-Square dissimilarity between them. The score level fusion is performed to provide an enhanced security system. A threshold value is kept and it is compared with the scores to authenticate the person. The Minimum Distance Classifier is used to identify the person by fusing the multimodal features.

Kinect-based human gait identifcation under different covariate factors

Introduction: Nowadays human gait identification/recognition is available in a variety of applications due to rapid advances in biometrics technology. This makes them easier to use for security and surveillance. Due to the rise in terrorist attacks during the last ten years research has focused on the biometric traits in these applications and they are now capable of recognising human beings from a distance. The main reason for my research interest in Gait biometrics is because it is unobtrusive and requires lower image/video quality compared to other biometric traits. Materials and Methods: In this paper we propose investigating Kinect-based gait recognition using non-standard gait sequences. This study examines different scenarios to highlight the challenges of non-standard gait sequences. Gait signatures are extracted from the 20 joint points of the human body using a Microsoft Kinect sensor. Results and Discussion: This feature is constructed by calculating the distances between each two joint points from the 20 joint points of the human body provided which is known as the Euclidean Distance Feature (EDF). The experiments are based on five scenarios, and a Linear Discriminant Classifier (LDC) is used to test the performance of the proposed method. Conclusions: The results of the experiments indicate that the proposed method outperforms previous work in all scenarios.