Human Monitoring Scanner Using GAIT

doi:10.21203/rs.3.rs-2088072/v1

Download PDF

Research Article

Human Monitoring Scanner Using GAIT

https://doi.org/10.21203/rs.3.rs-2088072/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Biometric system is useful to identify person. Biometric systems are rising and generally used technology. Because of this generalization today many protected systems are taking advantage of biometric systems. And one of these newest research areas of biometric systems is gait analysis systems. Numerous locations, including banks, airports, courtrooms for criminal proceedings, and security reasons, require the biometric technology. We can recognize a person by their fingerprints, handprints, eyes (retina), and face with the use of a biometric system . The issue of walking behavior is addressed through the use of gait technology. In security and video surveillance applications, gait identification is crucial. A person's lack of walking style may be resolved by reorganizing the gait. Gait biometric skill recognizes a individual by the walking pattern. Gait study investigates human body parts to find personal features for human recognition. Gait technology recognizes person walking pattern for individual detection. When person identification ratio matches up to 60% to 70% in order to that here Re-identification is solution. Several studies on this study area are examined during the survey.

Gait Analysis

Gait Recognition

Biometric systems

Human Identification

Gait biometric technology etc

In the world of arising terrorism and free movement of criminals, people evidently know the significance of safety monitoring and manage for the purposes of national justification and community security. The increasing demand for biometric features based safety systems demonstrated by the IT companies’ shows that this area will be one of the top study areas in the future. In fact some techniques in this discipline like fingerprint, Iris and face recognition is old formed. Early-warning systems by now have a great number of observation cameras installed in civic areas, but need smart approaches to person detection. A completely helpful structure would examine the composed record data and make public an attentive by an unfavorable incident happens. Detecting an irregular performance, the structure could immediately recognize all scene participants, quickly consider their earlier actions, and begin tracking the suspects. Gait discovery has more than a few rewards other biometric process. It doesn’t necessitate subject as human being in make in front of with it. By the way some other approaches like gait examination or 3D face detection becomes appear in the near future.

From 1995 to till date technical society has exposed an important attention in person recognition based on gait analysis. Day by day the technological improvements in hardware sector like high processing capacities made this study area feasible for biometric secure systems. Some papers regarding the gait analysis have been presented so far. For Example; [1] focused on major three modules: (1) 3D corpse pretense, form and presentation facts opinion system (2) gait semantic factor folding model; (3) gait semantic attribute distillation arrangement. [2] Designed a novel structure that is covariate aware to accommodate the presence of such covariates. The typical power shadow picture, a single 2-D spatiotemporal pattern from a film series, is used to describe how the human body moves during gait (AESI). [3] The active mark and fixed aspect, which are coefficients of the quadrature copy filter (QMF)-graph wavelet sort bank, are removed using the traverse wavelet convert and bipartite graph model, respectively. Normalization is followed by attribute mixture. [4] A robust, efficient, and gait-related loss utility called Angle Center Loss (ACL) is expected to be aware of discriminative gait facial appearance. [5] Multi-task generative adversarial networks (MGANs) are projected for exploring particular characteristics of knowledge views. In contemplation of save new chronological data, It also advise a fresh multi-channel gait pattern, known as period energy image (PEI). [6] in order to categories the abnormal gaits, the GRU classifier was constructed. Its performance was evaluated against that of other machine learning-based classifiers. [7] A discriminative projection and a new gait quality, GII, were jointly estimated for cross-view gait identification. The use of the cosine parallel dimension allows GII to be launched in agreement with minor analytical variations. [8] Expected a gait recognition algorithm based on a character-adaptive unseen Markov model (SAHMM). It first built an examination vector set and presented a quality mining approach based on a limited gait energy image (LGEI). This set is used to optimize the gait detection algorithm's SAHMM-based process parameters. In conclusion, the predicted method for gait detection under various conditions, such as cross views, person dressing, or sack carrying, etc., is thoroughly assessed based on the CASIA Dataset B. [9] A postured-based gait detection, an unique model-based system for identifying gaits. Both posture-based characteristics and posture-based categorization make up this concept. [10] Gait parameter mining, the origin and agreement of gait features, and the Kinect skeletal data stream are all planned. [11] This study found that a person's three-channel most stable excessive value region can enhance the number of recovered characteristic regions while performing attributes mining. [12] The exercise of an ANN-based gait signal analysis block for approximate gait signals on the chest from those captured by damaged sensors on the other body situation has been in the future, and a crucial improvement on the demonstration of the planned safety plan has been demonstrated from the trial results. [13] The JRC-CNN gait identification system is a new model-based gait detection system that is proposed in this paper. There are also some additional papers in gait examination focused on health situation like patients or older people at risk of falling. Improved variability of gait is related with amplified hazard of falling but in one paper explore in added parameters shows that limited active stability of gait can also count as a analyst of collapse danger. For the purpose of detecting humans by gait, Sundaresan projected a hidden Markov models (HMMs)-based structure. Huang extends the pattern identical approach to gait identification by fusing a revolution based on canonical analysis and an Eigen space revolution for feature selection. Kshitiz Varma and Sanjeev Sharma (2015) used front view figure data for gait recognition using PCA techniques. Worapan Kusakunniran, Qiang Wu, and Jian Zhang (2013) propose a novel structure to build a new view-invariant feature for cross-view gait recognition. A number of the normalized gait silhouettes are planned for and put to use in a Procuresses Shape Analysis (PSA) analysis. It can be used to remove a brand-new view-invariant gait feature. Only averaging the gait silhouettes above the gait cycle results in the GEI. Gait entropy picture, gait flow image, and chrono-gait image are additional appearance-based aspect characterizations (GEnI). The CGI is a sequential template in which the gait frames and temporal data are fixed using a colour mapping job. The color-coded gait form photos are assembled into a gait cycle to create the CGI. The GFI is produced by collective grey scale contours that span more than one gait cycle and visual flow. The GEnI stands for the pixel value uncertainty in the series of silhouette images. In total, a number of gait feature representations, such as the enhanced gait energy image (EGEI) and the covered gait energy image, are also based on the GEI (MGEI). In order to enhance the GEI attribute, the EGEI keeps the active region. The MGEI is constructed utilising the GEI and the entropy of each pixel to create a binary attribute choice GEI mask. The GEI is more helpful in gait detection than new gait quality evaluation, according to Iwama et al. Features like clothing, carrying goods, different outlooks, and most significantly, occlusion, increase the presentation of gait identification. Numerous shows for invariant gait attributes have been expanded, especially for the creation of invariant gait aspects for observation. Deep learning-based methods are quite good at identifying gait in almost every image and video they are applied to. Deep learning has recently gained widespread attention for its use in extracting beneficial and view-invariant gait features. For instance, a method for extracting invariant gait features using convolution neural networks (CNNs) has been made possible in. The authors used gait data, including GEIs, under various carrying conditions and varied viewing angles. With the aid of multi-stacked auto-encoders, developers can learn these traits in a broader sense. A CNN called GEINet is recommended by Shiraga et al. to extract view-invariant gait characteristics. Liu et al. use the roaming expressed person detection technique to extract 2D positions of joints as a storage-based detection system. After that, gait recognition is accomplished using a recurrent neural network (RNN). To recognise people and forecast various characteristics of their walking habits, Chao Yan et al. suggest employing a CNN and a multi-task learning model (MLT). Simultaneously Munif Alotaibi et.al elaborates a gait identification technique with CNN. They introduced a particular deep CNN framework with numerous convolutional and sub-sampled layers. This design is unaware of the previously mentioned covariant elements that influence the reduction of the detection ratio. Gait parameters are calculated using all sophisticated techniques using one complete gait cycle. Researchers claim that no technique has yet been able to recognize how difficult it is to detect motion from a limited gait cycle. Deep neural networks' fully convolution network (FCN) types are able to carry out frame-to-frame adjustments. It works well for object recognition and great declaration semantic segmentation in images and videos. A FCN for the statement "one image great" was introduced by Dong et al. This research may identify an end-to-end mapping between a low resolution image and a high resolution image, outperforming the outdated coding-based, specific image super-resolution technique. Our projected shape is based on FCN structural design for defective GEI to whole GEI conversion.

Law implementation and dissimilar establishment require a latest proficient method to path and re-recognize a person of notice through dissimilar cameras. In most cases, the subject of the attention is not named, and the distinctive video may be small and lack distinction. In order to discover and re-identify a stranger, the approach described in this research uses a novel regional-LSTM knowledge model and a 12-second walk. By developing a split LSTM model for each region, the proposed method initially focused on the mobility of events in various body regions. Results were then gathered from 22 locations in a sub network to remove transactions and various levels of individuality from all components. The mentioned regional LSTM model produced a gait-embedded vector to represent a 2-second walk. The findings demonstrate that, when applied to high and neutral datasets, the newly presented regional LSTM model outperforms existing techniques in terms of top-k accuracy, receiver operating characteristic (ROC) curves, and precision-recall (PR) curves. This shows that the suggested method performs well (CMC test), can expertly separate a matter's behaviors from those of other topics (ROC test), and is tenanted to be more relevant (PR test). The investigation's findings indicate that it is possible that one out of every four videos computed using the methods provided suggested the subject with accuracy levels of above 90.8% and 85.7% in both extreme and reasonable data, respectively. This experiments that the planned regional LSTM model is capable and helpful in tracing and re-recognize human being of importance.

The gait-connected loss task was directed using spatial-temporal parameters by Yuqi Zhang and Yongzhen Huang (2020). A cultured panel divides the shadow into 4 parallel sections. Each parallel fraction is divided up into a separate CNN. The frame-level CNN is used to generate consideration scores using the LSTM concentration model.

In 2020, Kooksug Jun, Yongwoo Lee, Sanghyub Lee, Deok-Won Lee, and Mun Sang Kim proposed a classifier using 3D framework data and gated recurrent units (GRUs). To characterize the abnormal gaits, they developed the GRU classifier and compared its performance to other machine learning-based classifiers.

Expanding a new Sahmm-based model for each personality in a specific gait dataset, Xiuhui Wang, Shiling Feng, and Wei Qi Yan (2019) prepared them utilizing a revolutionary gait illustration, called LGEI (local gait energy image). Yiwei He, Junping Zhang, Hongming Shan, and Liang Wang suggest using multi-task generative adversarial networks (MGANs) to render attributes that are determined by the knowledge perspective. They also suggest a novel multi-channel gait pattern called as period energy image to safeguard additional temporal data (PEI). Based on the presumption that diverse points of view exist, the MGANs can use adversarial preparation to extract additional discriminative parameters from gait sequences. Using the Kinect skeletal data stream, gait parameters extraction, gait features extraction, and classification, Muqing Deng and Cong Wang (2019) planned a technique in four stages. According to JIAN LUO 1 and TARDI TJAHJADI (2020), the gait identification method consists of three modules: the gait semantic constraint folding model, the gait semantic attribute refinement network, and the 3D body pose, figure, and presentation data inference complex (3D-BPSVeNet).

New covariate aware structures, according to Himanshu Aggarwal and Dinesh Kumar Vishwakarma (2018), were expected to contract as these covariates appeared. They do this by creating a single 2-D spatiotemporal pattern from a collection of videos, which they refer to as the common energy profile image (AESI).

In order to obtain the coefficients of the quadrature mirror filter (QMF)-graph wavelet filter bank, Sagar Arun More and Pramod Jagan Deore employed the cross wavelet transform to extract the dynamic feature and the bipartite graph model to extract the static aspect.

In cross-view gait identification, which has been advanced by introducing a modification expression to mechanically detain the primary discriminative data, a discriminative projection with list-wise constraints (DPLC) is anticipated to compact with vision variation. This is according to Zhaoxiang Zhang, IEEE, Jiaxin Chen, Qiang Wu, and ndLing Shao.

A novel gait identification system was created by three researchers, Nirattaya Khamsemanan, Cholwich Nattee, and Nitchan Jianwattanapaisarn, using a posture-based model. Both posture-based characteristics and posture-based classification make up this concept.

MIMI ZHOU presents a better approach using the nearly all stable three-channel extreme areas in order to retrieve the three-channel most stable extreme region. By using gait indication power changes and the development of an artificial neural network, Yingnan Sun, Benny Lo, and IEEE developed a novel (Body Sensor Network) BCS move toward secure wireless associations for wearable and implantable healthcare approach.

Gait-related loss tasks can be directed in a way that allows for the extraction of spatial-temporal data. A learnt partition divides the shadow into 4 horizontal sections. An individual CNN receives each horizontal component. On the frame-level CNN, production awareness scores are calculated using the LSTM concentration model.

Review of loss functions:

$$Ls=- \sum _{i=1}^{m}\text{log}\frac{{e}^{w\frac{T}{yi}xi+byi}}{{\sum }_{j=1}^{n}{e}^{w}{}_{j}{}^{T}xi+bj}$$

Where d is the attribute aspect and xi represents the i-th deep aspect of the y i-th class. Wj Rd, or the j-th column of weights, stands for the last entirely linked stratum, and b Rd stands for the prejudicial time. The class integer and the batch extent are, respectively, m and n. It has been demonstrated that the erudite parameters by the softmax loss are just sovereign rather than discriminative in face detection. Gaited Recurrent Neural Network: This technique initially assembles skeletal data by specifically employing Kinect v2 and Microsoft SDK to collect the 3D coordinates of 25 joints. Also created Kinect coordination using sic sensors as well as calibrated each sensor. After data collected GRU classifier useful on collected data. For the classification of abnormal gaits based on skeletons, a multilayered GRU classifier is used. After reaching the final stage of classification, the GRU classifier was able to differentiate between normal and abnormal gait utilizing information from skeletal gait. A particular dataset contained 80 to 90 frames of skeleton data. First 10 frames couldn’t used because of noise. Next 50 frames used. At least 60 frames essential for classification. The classifier is trained using the cross entropy task L2 regularization technique during the preparation phase.

L(x,y) = $\text{-}\sum _{\text{i=1}}^{\text{6}}\text{yi log(softmax(yi))+}\frac{\gamma }{2}$${\left|w\right|}^{2}$

Where L(x,y) and W represent, respectively, the cost related to the input data x and vector label y. Here cross substantiation method is used.

Multitask Generative adversarial Networks: As a new method of data distribution, generative adversarial networks were introduced. In this paper first introduced cross view gait detection then in Period Energy Image the methods are 1) Template building: It constructs a periodic signal from shadow image. A gait series' frames are also allocated to several channels in PEI. The following are typical outline images that represent a range of amplitude T(k).:

${PEI}_{k}$ = $\frac{1}{Nk}$ $\sum _{rt\in T\left(k\right)}Bt$,

Where

T (k) =$\left[\frac{k}{nc+1}- \frac{m}{2} , \frac{k}{nc+1}+ \frac{m}{2}\right]$.

m = Sliding window

${\text{N}}_{\text{k}}$ = silhouette image

2) Spatial and Computational Complexities: Gait templates' elevation and width are denoted by the letters h and w. The length of a full progression is Nk and the number of frames in the kth channel is N. There are n c channels in total.

Multi-Task Generative Adversarial Network consists of five components. 1) Encoder: The encoder, a convolutional neural network, provides a thorough phase for detection. 2) Angle of view Classifier: There are two softmax layers and two entirely related layers in the classifier. The classifier takes the view detailed parameters as the gait pattern as input. 3) View Transform Layer: The formulation for changing the view from angle u to v is

${\text{z}}^{\text{v}}$ = ${\text{z}}^{\text{u}}$ + $\sum _{\text{i}=\text{u}}^{\text{v}-1}\text{h}\text{i}$

Where, hi is change Vector.

3) Generator: The outcomes of view exchange fed into the initiator. The creator generates PEI. 5) The produced gait image x^v serves as the discriminator's input. Deterministic Learning & Data Stream of Microsoft kinect: planned methods have four parts. 1) Kinect Skeleton Data Stream: This technique is able of extracting RGB image. Only the skeleton data stream from the Kinect v1.0 is used in this technique. 2) Spatial-temporal features and kinematic parameters are the two groups that person gait techniques divide into. Spatial temporal features contain step length, footstep length shape width and so on. Joint angles between body pieces and joint activity during the gait cycle serve as indicators of kinematic characteristics.3) Gait Features Extraction: Gait parameters are used to establish the essential gait technology of the walking method as gait parameters which can characterize chronological changes of body arrangement and shape information.

4) Gait Recognition scheme: First step is training stage second comparing test sequences. Computer Vision analyzed and categorized human gait: 1) video to Frame Conversion: This steps proposed moving object recognition with video and it transformed into frames. 2) Moving object revealing: This step introduced two models in each picture. One for subject forepart and another for background. Original image and background image can changed into gray scale image.

Foreground Image = abs (bk - p)

Where abs = Absolute value

Bk = Background gray scale

P = Person gray scale

3) Image Conversion:

5) Proposed Features Extraction:

Figure shows the statistical extraction of gait features. From the SURF (Speed up robust features) descriptor, For instance, if there are n data points (x(i)) and k clusters, intention reason Z can be expressed as follows: Gait occasion discovery: The traditional k-means method was applied to the gait segment. If there are n data points and k clusters, for instance, intention reason Z can be expressed as:

Z= $\sum _{i=1}^{n}\sum _{j=1}^{k}wij$|${x}^{i }- {\mu }_{k}$|2

Where,

${\mu }_{k}=centroid of {x}^{i }$ ‘s cluster

Wij = 1 Otherwise wij – 0

The shape model demonstrated that the best k-means performance happens when the gait segment's chosen granularity level is two and is used for two clusters. Characteristics Body Model with practical clothing: The numerical & 3D mesh vertices are based on the statistical knowledge technique, and the 3D body dataset is formed of the deformation correlations among semantic body attributes. Based on silhouettes variation this paper projected 2D-3D-BPSDs judgment methods. The evaluation correctness still needs to develop appropriate lack of RGB data. Using multiple body outlines and position factors, 3D parametric body sculpts may generate a variety of 3D bodies. The view among body and camera continuously substitute when a subject is walking from remote detachment. In all gait detection these substitute are unnoticed.

There are various types of datasets in gait study world. Standard dataset, Multi view dataset, infrared gait dataset are main type of datasets. CASIA in one of the most important dataset which is shown in figure.

[1] The occlusion, carrying, and garment deviation datasets from the CMU MoBo, CASIA B, and KY4D. [2] The USF Human-ID gait challenge dataset, the OU-ISIR treadmill dataset, and the CASIA Dataset B are among the datasets that are readily available. For categorization, linear kernel SVMs are used, with penalty parameter c = 1 and tolerance for completion criteria = 1 e 4. In the example, there are 124 individuals in the multiview dataset CASIA dataset B. [4] In the work, CASIA B multi-view gait database, this contains 124 subjects. Each person is displayed in 10 series with various variables (CVs), such as slow walking (nm-01 to nm-06), carrying a bag (bg-01, bg-02), and wearing apparel (cv-01, cv-03) (cl-01, cl- 02). Figure depicts the intended procedure. [5] To determine the best settings, this research will first evaluate the contact of each gait identification characteristic. A number of cross-view gait detection datasets, including CASIA-B, OULP, and OUMVLP, are used to test the suggested methodology. These three datasets all provide silhouettes, which can help with more accurate temporal modeling. To better comprehend the qualitative causes of distinct defeat functions, we take into account the learned gait characteristics. [6] In our trials, three commonly used gait datasets—OU-ISIR, CASIA-B, and USF—are used. Figure 3 shows a few samples from each of the three datasets. The OU-ISIR gait dataset has a sizable population. A popular cross-view gait dataset is CASIA-B. It can hold 124 different subjects and has 11 additional view angles, including 0, 18, 36, 54, 72, 90, 108, 126, 144, 162, and 180 degrees. USF is a normally used gait database too. In total, USF offers 122 different subjects. Every subject's gait series is sampled below the special effects of five variables, which is closer to a logical progression.

One method for recognising people is called "gait recognition and re-identification" [43].Re-reorganization is the process of comparing images or videos of the same person taken at different angles, times, and with different cameras. A split LSTM network is creating to remove significant details from chronological information of every region of body. Re-identification Test: 1) Cumulative Matching Characteristics (CMC) 2) Receiver Operational Features (ROC) 3) The Precision Recall Curve (PR) LSTM technique is better suited for recognition than the GEI approach. [28] 12 second walk recorded at a distance by the Kinect V1 device. 240 successive skeleton frames are present throughout a 12 second stroll. Regional representation: In Table, a region is defined as three linked joints in a typical frame. There are joints at the left, middle, and right of each region. With a focus on region r of the normal frame, I define Ci r Left as a vector that starts at the centre joint and ends at the left joint. Gait detection is a type of behavioral biometric verification that identifies and confirms people by their walking manner and speed. Identification of a human being by a walking patternBy building an individual LSTM model component for each location, the initial targets are patterns of action in diverse bodily regions. A gait-surrounded vector is created by the LSTM model to represent a person walking. Model based approach utilizes body actions (Dynamic features) like pace lengths and limb travels through body structures as gait parameters. Standard deviations and averages of arm angle, pace length, and momentum were used as dynamic characteristics.

In this exertion, a survey on human detection based on gait analysis has been accessible. By combining statistical data from the SURF and Haralick parameters, current acceptance statistical gait distinctiveness can identify the biometric gait characteristic for person recognition. The accuracy of person detection can be increased by using these characteristics to identify persons and to reduce the impact of "covariate variables ".By the way, the reported works show encouraging results in routine gait analysis, but further architectures need to be implemented and experienced. In particular, for gait analysis, the most successful way seems to consist in the arrangement of Deep Learning with classical Machine learning models. Moreover comparing the performance will be interesting in the future.

In the paper, I suggested a novel gait detection and re-reorganization capacity using a projected regional LSTM learning model as the basis. The conscious objective is to maintain a brief, freestyle 2-second saunter. It is built on the principle that every part of the body has its own unique periodic faction when moving, and that certain of the activities of some parts of the body may have more unique distinctiveness than others. To separate the important information from the chronological data of each body area, a split LSTM model is built. A area of the body acquires its unique character through a process that preserves the regional growth pattern.

Title: Human Monitoring Scanner Using GAIT

Name of First Author: Kiran Macwan

E-mail address of First Author: [email protected]

Name of Corresponding Author: Dr. Harshal Shah

E-mail address of Corresponding Author: [email protected]

Ethical Approval:

I hereby certify Kiran Macwan and Dr. Harshal Shah of the Human Monitoring Scanner using GAIT that the following is true:

1) This material is the author's own original work and has not been previously published elsewhere.

2) This paper is not currently being considered for other publications.

3) The article truthfully and completely reflects the author's own research and analysis.

4) This paper duly acknowledges the significant contributions of co-authors and collaborators.

5) Findings are well placed in the context of previous and existing research. 6) All sources used are properly cited (correct citation). A verbatim copy of the text should be marked with quotation marks and the source.

7) All authors are personally and actively involved in the substantive work that led to this publication and accept public responsibility for their content.

Competing interests

Please be directly or indirectly related to the submitted work. The report must be submitted within the past three years from the start of work (conducting the study and preparing the work for submission). Disclosure of Interests provides a complete and transparent process to help readers make their own judgments about possible bias. This does not mean that any financial relationship with the organization that sponsored the research or remuneration for consulting services is inappropriate. In this manuscript, I agree to disclose any potential financial or other conflicts of interest that could be perceived as affecting the objectivity of my research.

Authors' contributions

The final draught of the manuscript has been read and approved.

We have made significant contributions to the work that has been presented, including study design, data collection, analysis, and interpretation.

We have significantly aided in the creation of the manuscript and/or provided major intellectual content changes.

We shall be responsible for every element of the work that is submitted.

We consent to assist with any inquiries and/or investigations into the veracity and integrity of the submitted work.

Funding: Not Applicable

Availability of data and materials: Not Applicable

Y. Zhang, Y. Huang, S. Yu and L. Wang, "Cross-View Gait Recognition by Discriminative Feature Learning," in IEEE Transactions on Image Processing, vol. 29, pp. 1001-1015, 2020, doi: 10.1109/TIP.2019.2926208.
K. Jun, Y. Lee, S. Lee, D. -W. Lee and M. S. Kim, "Pathological Gait Classification Using Kinect v2 and Gated Recurrent Neural Networks," in IEEE Access, vol. 8, pp. 139881-139891, 2020, doi: 10.1109/ACCESS.2020.3013029
X. Wang, S. Feng and W. Q. Yan, "Human Gait Recognition Based on Self-adaptive Hidden Markov Model," in IEEE/ACM Transactions on Computational Biology and Bioinformatics, doi: 10.1109/TCBB.2019.2951146.
Y. He, J. Zhang, H. Shan and L. Wang, "Multi-Task GANs for View-Specific Feature Learning in Gait Recognition," in IEEE Transactions on Information Forensics and Security, vol. 14, no. 1, pp. 102-113, Jan. 2019, doi: 10.1109/TIFS.2018.2844819.
M. Deng and C. Wang, "Human Gait Recognition Based on Deterministic Learning and Data Stream of Microsoft Kinect," in IEEE Transactions on Circuits and Systems for Video Technology, vol. 29, no. 12, pp. 3636-3645, Dec. 2019, doi: 10.1109/TCSVT.2018.2883449
Elharrouss, O., Almaadeed, N., Al-Maadeed, S. et al. Gait recognition for person re-identification. J Supercomput 77, 3653–3672 (2021).
K. Z. C. Tun and S. M. M. Zaw, "Gait based Human Identification through Intra-Class Variations," 2021 6th International Conference on Inventive Computation Technologies (ICICT), 2021, pp. 564-571.
Fangmin Sun a , Weilin Zang a , Raffaele Gravina b , Giancarlo Fortino b , Ye Li et al. “Gait-based identification for elderly users in wearable healthcare systems” 2020 ELSEVIER Journals Volume 53, Pages 134-144.
S. Chakraborty and A. Nandy, "An Unsupervised Approach For Gait Phase Detection," 2020 4th International Conference on Computational Intelligence and Networks (CINE), 2020, pp. 1-5.
X. Shao, X. Nie, X. Zhao, R. Zheng and D. Guo, "Gait Recognition based on Improved LeNet Network," 2020 IEEE 4th Information Technology, Networking, Electronic and Automation Control Conference (ITNEC), 2020, pp. 1534-1538.
L. Gong, J. Li, M. Yu, M. Zhu and R. Clifford, "A novel computer vision based gait analysis technique for normal and Parkinson’s gaits classification," 2020 IEEE Intl Conf on Dependable, Autonomic and Secure Computing, Intl Conf on Pervasive Intelligence and Computing, Intl Conf on Cloud and Big Data Computing, Intl Conf on Cyber Science and Technology Congress (DASC/PiCom/CBDCom/CyberSciTech), 2020, pp. 209-215.
J. Kondragunta and G. Hirtz, "Gait Parameter Estimation of Elderly People using 3D Human Pose Estimation in Early Detection of Dementia," 2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC), 2020, pp. 5798-5801.
T. Jiang, Y. Ge and W. Li, "Gait analysis of people relying on mobility aids by using laser range finder," 2020 IEEE International Conference on Human-Machine Systems (ICHMS), 2020, pp. 1-4.
G. Pagano, G. D’Addio, M. De Campi, L. Donisi, A. Biancardi and M. Cesarelli, "Rehabilitation Outcome in Patients undergone Hip or Knee Replacement Surgery using Inertial Technology for Gait Analysis," 2020 IEEE International Symposium on Medical Measurements and Applications (MeMeA), 2020, pp. 1-5.
J. Luo and T. Tjahjadi, "View and Clothing Invariant Gait Recognition via 3D Human Semantic Folding," in IEEE Access, vol. 8, pp. 100365-100383, 2020.
P. Limcharoen, N. Khamsemanan and C. Nattee, "View-Independent Gait Recognition Using Joint Replacement Coordinates (JRCs) and Convolutional Neural Network," in IEEE Transactions on Information Forensics and Security, vol. 15, pp. 3430-3442, 2020.
M. Zhou, "Feature Extraction of Human Motion Video Based on Virtual Reality Technology," in IEEE Access, vol. 8, pp. 155563-155575, 2020
ATHIRA NAMBIAR, ALEXANDRE BERNARDINO, and JACINTO C. NASCIMENTO, ACM Computing Surveys Vol.52 April 2019 Article No.: 33.
H. Chiu, C. Lin and T. S. Li, "Gait Recognition using Histogram of Oriented Gradient and Self-Organizing Feature Map Classification in Variable Walking Speed," 2019 International Conference on Fuzzy Theory and Its Applications (iFUZZY), 2019, pp. 283-286.
P. P. Min, S. Sayeed and T. S. Ong, "Gait Recognition Using Deep Convolutional Features," 2019 7th International Conference on Information and Communication Technology (ICoICT), 2019, pp. 1-5.
Macoveciuc I, Rando CJ, Borrion H. Forensic Gait Analysis and Recognition: Standards of Evidence Admissibility. J Forensic Sci. 2019 Sep;64(5):1294-1303.
R. G. Birdal, A. Sertbaş and B. Mïhendisliği, "Human Identification Based on Gait Analysis: A survey," 2018 3rd International Conference on Computer Science and Engineering (UBMK), 2018, pp. 489-493.
Peibei Cao, Weijie Xia , Ming Ye, Jutong Zhang, Jianjiang Zhou “Radar-ID: human identification based on radar micro-Doppler signatures using deep convolutional neural networks” Volume 12, Issue 7, July 2018, p. 729 – 734.
Imad Rida, Noor Almaadeed, Somaya Almaadeed “Robust gait recognition: a comprehensive survey” IET Biometrics ,August 2018.
MICHAŁOWSKA,WALCZAK,GRABSKI,CIEŚLAK “People Identification Based on Dynamic Determinants of Human Gait” Vibrations in Physical Systems 2018,Vol. 29,pp 1-6.
K. Z. Htun and S. Maung Maung Zaw, "Human Identification System Based on Statistical Gait Features," 2018 IEEE/ACIS 17th International Conference on Computer and Information Science (ICIS), 2018, pp. 508-512.
Lubov Shiripova , Evgeny Myasnikov “Comparative Analysis of Classification Methods for Human Identification by Gait” Samara University, Moskovskoe Shosse 34, Vol 4 pp 51-59.
Maryam Babaeea,, Linwei Lia, Gerhard Rigolla “Person Identification from Partial Gait Cycle Using Fully Convolutional Neural Network” a Institute for Human-Machine Communication April,18 .
Albert Haque, Alexandre Alahi, Li Fei-Fei “Recurrent Attention Models for Depth-Based Person Identification” arXiv.org V 1 ,2018.
N. Khamsemanan, C. Nattee and N. Jianwattanapaisarn, "Human Identification From Freestyle Walks Using Posture-Based Gait Feature," in IEEE Transactions on Information Forensics and Security, vol. 13, no. 1, pp. 119-128, Jan. 2018.
H. Aggarwal and D. K. Vishwakarma, "Covariate Conscious Approach for Gait Recognition Based Upon Zernike Moment Invariants," in IEEE Transactions on Cognitive and Developmental Systems, vol. 10, no. 2, pp. 397-407, June 2018.
Z. Zhang, J. Chen, Q. Wu and L. Shao, "GII Representation-Based Cross-View Gait Recognition by Discriminative Projection With List-Wise Constraints," in IEEE Transactions on Cybernetics, vol. 48, no. 10, pp. 2935-2947, Oct. 2018.
S. A. More and P. J. Deore, "Gait Recognition by Cross Wavelet Transform and Graph Model," in IEEE/CAA Journal of Automatica Sinica, vol. 5, no. 3, pp. 718-726, May 2018.
Wu Liu, Cheng Zhang, Huadong Ma, Shuangqun Li “Learning Efficient Spatial-Temporal Gait Features with Deep Learning for Human Identification” Neuroinformatics . 2018 Oct;16(3-4):457-471.
Sungjun Hong and Euntai Kim “A New Automatic Gait Cycle Partitioning Method and Its Application to Human Identification” Int. J. Fuzzy Log. Intell. Syst. 2017;17(2):51-57
S. Yu, H. Chen, E. B. G. Reyes and N. Poh, "GaitGAN: Invariant Gait Feature Extraction Using Generative Adversarial Networks," 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), 2017, pp. 532-539.
Liao R., Cao C., Garcia E.B., Yu S., Huang Y. (2017) Pose-Based Temporal-Spatial Network (PTSN) for Gait Recognition with Carrying and Clothing Variations. In: Zhou J. et al. (eds) Biometric Recognition. CCBR 2017. Lecture Notes in Computer Science, vol 10568. Springer, Cham.
Lishani, A.O., Boubchir, L., Khalifa, E. et al. Human gait recognition based on Haralick features. SIViP 11, 1123–1130 (2017).
Manjunatha Guru, Kamalesh , Dinesh , July 2017 International Journal of Image, Graphics and Signal Processing 9(7):45-54.
G. Mokhtari, Q. Zhang, C. Hargrave and J. C. Ralston, "Non-Wearable UWB Sensor for Human Identification in Smart Home," in IEEE Sensors Journal, vol. 17, no. 11, pp. 3332-3340, 1 June1, 2017.
X. Chen, X. Luo, J. Weng, W. Luo, H. Li and Q. Tian, "Multi-View Gait Image Generation for Cross-View Gait Recognition," in IEEE Transactions on Image Processing, vol. 30, pp. 3041-3055, 2021, doi: 10.1109/TIP.2021.3055936.
Wareechol and W. Chiracharit, "Recognition of Similar Gait Pattern Using Transfer Learning DarkNet," 2021 9th International Electrical Engineering Congress (iEECON), 2021, pp. 381-384, doi: 10.1109/iEECON51072.2021.9440386.
P. Limcharoen, N. Khamsemanan and C. Nattee, "Gait Recognition and Re-Identification Based on Regional LSTM for 2-Second Walks," in IEEE Access, vol. 9, pp. 112057-112068, 2021, doi: 10.1109/ACCESS.2021.3102936

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Human Monitoring Scanner Using GAIT

Status:

Version 1

Abstract

Figures

I. Introduction

Ii. Literature Surevy

Iii. Exsiting Methods

Iv. Dataset & Methods

V. Conclusion

Declarations

References

Additional Declarations

Status:

Version 1