ISO/IEC 39794-17:2021
(Main)Information technology — Extensible biometric data interchange formats — Part 17: Gait image sequence data
Information technology — Extensible biometric data interchange formats — Part 17: Gait image sequence data
This document specifies examples of application-specific requirements, recommendations and best practices in data acquisition applicable to gait image sequence data. Its typical applications include: a) support for human examination of high-resolution video and still images; b) support for human biometric verification and identification based on video and still images; c) automated gait image sequence verification and identification. This document ensures that image sequences are suitable for human identification and human verification generated by video surveillance and other similar systems. The following topics are not in scope of this document: — Definitions for facial and/or full body image related biometric profiles, which are fully covered in ISO/IEC 39794-5 and ISO/IEC 39794-16 respectively. — Security aspects like digital image sequence electronic signature, Presentation Attack Detection (PAD) and morphing prevention.
Titre manque — Partie 17: Titre manque
General Information
Relations
Standards Content (Sample)
INTERNATIONAL ISO/IEC
STANDARD 39794-17
First edition
2021-06
Information technology — Extensible
biometric data interchange formats —
Part 17:
Gait image sequence data
Reference number
©
ISO/IEC 2021
© ISO/IEC 2021
All rights reserved. Unless otherwise specified, or required in the context of its implementation, no part of this publication may
be reproduced or utilized otherwise in any form or by any means, electronic or mechanical, including photocopying, or posting
on the internet or an intranet, without prior written permission. Permission can be requested from either ISO at the address
below or ISO’s member body in the country of the requester.
ISO copyright office
CP 401 • Ch. de Blandonnet 8
CH-1214 Vernier, Geneva
Phone: +41 22 749 01 11
Email: copyright@iso.org
Website: www.iso.org
Published in Switzerland
ii © ISO/IEC 2021 – All rights reserved
Contents Page
Foreword .v
Introduction .vi
1 Scope . 1
2 Normative references . 1
3 Terms and definitions . 2
4 Abbreviated terms . 3
5 Conformance . 4
6 Modality specific information . 5
6.1 Purpose . 5
6.2 Practices . 5
6.3 Data models for gait recognition . 5
6.3.1 General. 5
6.3.2 Model-based methods . 6
6.3.3 Appearance-based methods . 6
6.4 Data flow of gait recognition . 7
6.5 Body tree concept for gait . 7
6.6 Camera image sequence requirements . 8
6.7 Gait recognition recordings . 9
6.7.1 General. 9
6.7.2 Gait and upper body movement image encoding . 9
6.7.3 Gait and upper body camera image resolutions . 9
6.8 Gait modality.10
6.8.1 General.10
6.8.2 Gait silhouette .10
6.8.3 Surveillance systems .12
6.9 Upper body movement modality .12
6.9.1 General.12
6.9.2 Face Features Motion (FFM) .13
6.9.3 Head movement recognition .13
6.9.4 Head Movements Static Body (HMS) .13
6.9.5 Head Movements Dynamic Body (HMD) .13
6.9.6 Hands movement .14
7 Profile-specific information .14
7.1 Purpose .14
7.1.1 General.14
7.1.2 Gait representations .14
7.1.3 Scene requirements . .14
7.2 2D gait image sequence profile .15
7.2.1 General.15
7.2.2 Gait image sequence representation profile requirements .15
7.2.3 Post-acquisition processing .15
7.2.4 Neural network training and testing .16
7.3 UBM 2D upper body movement profile .17
8 Encoding .17
8.1 Tagged binary encoding .17
8.2 XML encoding .17
9 Registered BDB format identifiers .18
Annex A (informative) Conditions for capturing .19
Annex B (informative) Encoding examples .30
© ISO/IEC 2021 – All rights reserved iii
Annex C (informative) Image sequence acquisition measurements .42
Bibliography .54
iv © ISO/IEC 2021 – All rights reserved
Foreword
ISO (the International Organization for Standardization) and IEC (the International Electrotechnical
Commission) form the specialized system for worldwide standardization. National bodies that are
members of ISO or IEC participate in the development of International Standards through technical
committees established by the respective organization to deal with particular fields of technical
activity. ISO and IEC technical committees collaborate in fields of mutual interest. Other international
organizations, governmental and non-governmental, in liaison with ISO and IEC, also take part in the
work.
The procedures used to develop this document and those intended for its further maintenance are
described in the ISO/IEC Directives, Part 1. In particular, the different approval criteria needed for
the different types of document should be noted. This document was drafted in accordance with the
editorial rules of the ISO/IEC Directives, Part 2 (see www .iso .org/ directives or www .iec .ch/ members
_experts/ refdocs).
Attention is drawn to the possibility that some of the elements of this document may be the subject
of patent rights. ISO and IEC shall not be held responsible for identifying any or all such patent
rights. Details of any patent rights identified during the development of the document will be in the
Introduction and/or on the ISO list of patent declarations received (see www .iso .org/ patents) or the IEC
list of patent declarations received (see patents.iec.ch).
Any trade name used in this document is information given for the convenience of users and does not
constitute an endorsement.
For an explanation of the voluntary nature of standards, the meaning of ISO specific terms and
expressions related to conformity assessment, as well as information about ISO's adherence to the
World Trade Organization (WTO) principles in the Technical Barriers to Trade (TBT) see www .iso .org/
iso/ foreword .html. In the IEC, see www .iec .ch/ understanding -standards.
This document was prepared by Joint Technical Committee ISO/IEC JTC 1, Information technology,
Subcommittee SC 37, Biometrics.
A list of all parts in the ISO/IEC 39794 series can be found on the ISO website.
Any feedback or questions on this document should be directed to the user’s national standards body. A
complete listing of these bodies can be found at www .iso .org/ members .html and www .iec .ch/ national
-committees.
The purchase of this ISO/IEC document carries a copyright licence for the purchaser to use ISO/IEC
copyright in the schemas in the annexes to this document for the purpose of developing, implementing,
installing and using software based on those schemas, subject to ISO/IEC licensing conditions set out in
the schemas.
© ISO/IEC 2021 – All rights reserved v
Introduction
Most countries around the world use biometric recognition systems for law enforcement and border
control. Many of these systems are not limited to face recognition purposes. To be consistent in such
deployments and processes, technical documents, guidelines and best practice recommendations
are being developed by different groups. However, these documents are primarily focused on travel
documents and related border control systems and the technical and operational issues to be considered
when planning and deploying them. Gait recognition is the biometric mode used as a secondary mode
in addition to biometric full body recognition or for forensic purposes. Face recognition is the biometric
mode best suited to the practicalities of travel documents and automated border processing.
There is little guidance covering the gait imaging for cross-border interoperability or law enforcement
services. There is a need for guidance for the use of high-quality digital cameras and video surveillance
devices to record gait image sequence data. This document is not restricted to full body gait image
sequence data. For example, it can be possible to extract only head movement data for recognition. Gait
recognition in this document therefore also covers recognition based on different body parts, e.g. head
or limb.
To enable applications on a wide variety of devices, including devices that have limited data storage,
and to improve biometric recognition accuracy, this document addresses not only data format, but also
scene constraints (lighting, pose, expression, etc.), photographic properties (positioning, camera focus,
etc.), and digital image attributes (image resolution, image size, etc.).
A specific biometric profile for cross-border interoperability is required for gait video and still images.
Gait image sequence data standardization is required to achieve the threshold quality gait image
database records required for automated gait biometric verification and identification. At the moment,
border guards record gait video using local practices for gait biometric enrolment, verification and
identification.
In order to fully understand the requirements implied in this document it is recommended that the user
become acquainted with the following documents: ISO/IEC 39794-16, specifying full body image file
formats; ISO 22311, giving information on a common output file format that can be extracted from the
video-surveillance contents collection systems to perform necessary processing; the ISO/IEC 30137
[7]
series, giving information on the use of biometrics in video surveillance systems; and EN 62676
defining video surveillance systems for use in security applications.
This document is intended to provide advice on the use of body image data for gait and upper body
movement recognition applications requiring exchange of gait image sequence data and upper body
movement data. Typical applications are:
— automated body biometric verification and identification (one-to-one as well as one-to-many
comparison),
— support for human biometric verification by comparison of persons based on video and still gait
images, and
— support for human examination of video and still gait images with sufficient resolution to allow a
human examiner to perform biometric verification.
The structure of the data format is compatible with ISO/IEC 39794-5 and ISO/IEC 39794-16.
This document specifies application-specific profiles including scene constraints, imaging properties
and digital image attributes, like image spatial and temporal sampling rates, image size, etc. These
modality and application profile specifics are contained in Figures 6 and 7 respectively. Data creation
and exchange is described in ISO/IEC 39794-16. The body image data blocks used in encoding gait
image sequence data are of type BodyImageDataBlockType, which is defined in ISO/IEC 39794-16. This
document makes normative reference to other ISO/IEC International Standards.
vi © ISO/IEC 2021 – All rights reserved
INTERNATIONAL STANDARD ISO/IEC 39794-17:2021(E)
Information technology — Extensible biometric data
interchange formats —
Part 17:
Gait image sequence data
1 Scope
This document specifies examples of application-specific requirements, recommendations and best
practices in data acquisition applicable to gait image sequence data. Its typical applications include:
a) support for human examination of high-resolution video and still images;
b) support for human biometric verification and identification based on video and still images;
c) automated gait image sequence verification and identification.
This document ensures that image sequences are suitable for human identification and human
verification generated by video surveillance and other similar systems.
The following topics are not in scope of this document:
— Definitions for facial and/or full body image related biometric profiles, which are fully covered in
ISO/IEC 39794-5 and ISO/IEC 39794-16 respectively.
— Security aspects like digital image sequence electronic signature, Presentation Attack Detection
(PAD) and morphing prevention.
2 Normative references
The following documents are referred to in the text in such a way that some or all of their content
constitutes requirements of this document. For dated references, only the edition cited applies. For
undated references, the latest edition of the referenced document (including any amendments) applies.
ISO/IEC 10918-1, Information technology — Digital compression and coding of continuous-tone still
images: Requirements and guidelines
ISO/IEC 10918-5, Information technology — Digital compression and coding of continuous-tone still
images: JPEG File Interchange Format (JFIF) — Part 5:
ISO/IEC 14496-1, Information technology — Coding of audio-visual objects — Part 1: Systems
ISO/IEC 14496-2, Information technology — Coding of audio-visual objects — Part 2: Visual
ISO/IEC 15444-1, Information technology — JPEG 2000 image coding system — Part 1: Core coding system
ISO/IEC 15948, Information technology — Computer graphics and image processing — Portable Network
Graphics (PNG): Functional specification
ISO/IEC 2382-37, Information technology — Vocabulary — Part 37: Biometrics
ISO/IEC 39794-1, Information technology — Extensible biometric data interchange formats — Part 1:
Framework
© ISO/IEC 2021 – All rights reserved 1
ISO/IEC 39794-5, Information technology — Extensible biometric data interchange formats — Part 5:
Face image data
ISO/IEC 39794-16, Information technology — Extensible biometric data interchange formats – Part 16:
Full body image data
XML Schema Part 0: Primer Second Edition, W3C Recommendation, October 2004, https:// www .w3
.org/ TR/ xmlschema -0/
XML Schema Part 1: Structures Second Edition, W3C Recommendation, 28 October 2004, http:// www
.w3 .org/ TR/ xmlschema -1/
XML Schema Part 2: Datatypes Second Edition, W3C Recommendation, 28 October 2004, http:// www
.w3 .org/ TR/ xmlschema -2/
3 Terms and definitions
For the purposes of this document, the terms and definitions given in ISO/IEC 39794-1, ISO/IEC 39794-16,
and ISO/IEC 2382-37 and the following apply.
ISO and IEC maintain terminological databases for use in standardization at the following addresses:
— ISO Online browsing platform: available at https:// www .iso .org/ obp
— IEC Electropedia: available at https:// www .electropedia .org/
3.1
3D model
DEPRECATED: 3D image
three-dimensional biometric capture subject body representation that encodes a surface or a volumetric
shape in a 3D space
Note 1 to entry: a 3D model can be a heavily processed biometric subject body 3D shape.
3.2
biometric profile
conforming subsets or combinations of base standards used to effect specific biometric functions
Note 1 to entry: Biometric profiles define specific values or conditions from the range of options described in
the relevant base standards, with the aim of supporting the interchange of data between applications and the
interoperability of systems.
[SOURCE: ISO/IEC 24713-1:2008, 3.9]
3.3
full body recognition
automated recognition of individuals based on their morphology
Note 1 to entry: This can include any or all of the head, torso and limbs.
3.4
gait recognition
automated recognition of individuals based on their manner of walking
3.5
human identification
process of searching through a list of biometric capture subject images to match against an input
image(s)
Note 1 to entry: Also known as one-to-many (1: N) searching.
2 © ISO/IEC 2021 – All rights reserved
3.6
vignetting
reduction of image brightness or saturation toward the periphery compared to the image centre
4 Abbreviated terms
AVC advanced video coding
BAP body animation parameter
BDB biometric data block
BER basic encoding rules
CCTV closed-circuit television
CEN European Committee for Standardization
CIE International Commission on Illumination (Commission
Internationale de l'Eclairage)
CNN convolutional neural network
DCI Digital Cinema Initiatives consortium
DCNN deep convolutional neural network
DER distinguished encoding rules
DL deep learning
EXIF exchangeable image file format
FAP face animation parameter
FFM face features motion
FOV field of view
GEI gait energy image
GHM gesture hand motion
HD high definition or horizontal deviation angle
HDR high dynamic range
HMD head movements dynamic body
HMS head movements static body
ICS implementation conformance statement
INTERPOL International Criminal Police Organization
ISO International Organization for Standardization
JFIF JPEG file interchange format
JPEG image compression standard specified as ISO/IEC 10918
© ISO/IEC 2021 – All rights reserved 3
JPEG2000 image compression standard specified as ISO/IEC 15444
JTC Joint Technical Committee
MP4 ISO/IEC 14496-14 digital multimedia file format used to
store video and audio
MPEG Moving Picture Experts Group
MPEG-4 ISO/IEC 14496-2 video compression format
MTF modulation transfer function
MTF20 highest spatial frequency where the MTF is 20 % or
above
NTSC National Television System Committee analogue televi-
sion colour system
PAD presentation attack detection
PNG portable network graphics format
RGB red green blue colour representation
SD standard-definition television
SFR spatial frequency response
THz terahertz
UBM2D upper body movement in 2D
UHD ultra-high definition
USAF US Air Force
VGA video graphics array image format having width 640
pixels and height 480 pixels
XML extensible markup language
XSD XML schema definition
5 Conformance
A BDB conforms to this document if it satisfies all relevant normative requirements related to:
— Its data structure, data values and the relationships between its data elements given in
ISO/IEC 39794-16.
— The relationship between its data values and the input biometric data from which the BDB was
generated as specified in ISO/IEC 39794-16.
— The application profile-specific conformance specifications given in Clause 8.
A system that produces BDBs is conformant to this document if all BDBs that it outputs conform to this
document (as defined above) as claimed in the ICS associated with that system. A system does not need
to be capable of producing BDBs that cover all possible aspects of this document, but only those that are
claimed to be supported by the system in the ICS.
4 © ISO/IEC 2021 – All rights reserved
A system that uses BDBs is conformant to this document if it can read, and use for the purpose intended
by that system, all BDBs that conform to this document (as defined above) as claimed in the ICS
associated with that system. A system does not need to be capable of using BDBs that cover all possible
aspects of this document, but only those that are claimed to be supported by the system in an ICS.
Conformity with this document also requires conformance with the record format specification defined
in ISO/IEC 39794-16.
6 Modality specific information
6.1 Purpose
This clause contains modality specific information, where a biometric modality is an information
category of a human trait. In general, there are various traits present in humans, which can be used as
biometric modalities. There are three human trait categories: the physiological, the behavioural and
the combination type of physiological and behavioural modalities. Gait and upper body movement are
behavioural modalities.
This clause also describes the requirements and best practice recommendations to be applied for gait
and upper body movement image sequence capturing in the application case of enrolment of biometric
reference data for feature databases. Conditions for capturing are discussed in more detail in Annex A.
6.2 Practices
The reliable extraction of characteristic features from image sequences and their recognition are
important issues in gait and upper body movement recognition. The basic body movement video or
a sequence of still images forms the basis for further analysis processing steps. Gait and upper body
movement are considered in this document to be the coordinated, cyclic combination of movements
that result in human locomotion.
For certain criteria, there may be two different levels: a minimum requirement and a best practice
recommendation. The wording is shown in Table 1. The requirement gives the minimum acceptable
values or value ranges in order to reach conformance. The best practice recommendation gives values
that result in better overall performance or quality, and users are encouraged to adopt best practice
values whenever possible.
Table 1 — Summary of wording for minimum requirements and best practice recommendations
Provision Wording
Requirement … shall …
Best Practice … should …
6.3 Data models for gait recognition
6.3.1 General
Gait recognition system can be classified depending on the sensors used in three groups, namely;
motion imaging (vision)-based, wearable sensor-based and spatial (floor) sensor-based. The motion
imaging (vision) can be divided into two groups, namely: appearance-based methods and model-based
methods. The appearance-based method can be also subdivided into two types; state space methods and
[9]
spatiotemporal methods . As stated in the Scope, this document is restricted to the motion imaging-
based gait recognition, which may use the whole available electro-magnetic spectrum available, not
only the visible bandwidth. The scope of this document is marked with bold text and continuous box
outline in Figure 1.
© ISO/IEC 2021 – All rights reserved 5
Figure 1 — Classification of gait recognition systems.
6.3.2 Model-based methods
Model-based approaches build a human body gait model and the extracted features of gait sequences
are fitted to that model. These methods are not sensitive to the individual’s appearance and clothing
but have high computational cost. It is hoped that the use of machine learning will enhance both the
creation of models and the least error model selection.
Model-based feature extraction is used to extract human joints (vertex positions). A vision-based
system for human motion analysis consists of three main phases: detection, tracking and perception. In
the last phase, a high-level description is produced based on the features extracted during the previous
phases from the temporal video stream. Marker-based solutions rely primarily on markers or sensors
attached at key locations of the human body.
Gait image sequence enrolment and identification using visual surveillance require the deployment of
an automated marker less vision system to extract the joints’ trajectories. Automated extraction of the
joints’ positions is a difficult task as non-rigid human motion encompasses a wide range of possible
motion transformations due to its highly flexible structure and to self-occlusion. Clothing type,
segmentation errors and different viewpoints pose a challenge for accurate joint localization. For a
model-based approach, a shape model is a priori established to match real images to this predefined
[10]
model, and thereby extract the corresponding features once the best match is obtained .
6.3.3 Appearance-based methods
Appearance-based methods or model-free gait recognition methods work directly on the gait
sequences. They do not use a model for the human body to rebuild human walking steps. These methods
have the advantage of low computational cost in comparison with model-based approaches, but the
disadvantages are sensitivity to changes in clothing and appearance. Applying an averaged silhouette
of a biometric subject during a gait cycle or using information obtained from a submillimetre image
enhances the silhouette image accuracy.
The decision to omit the state space from the scope of this document is based on the present status of
non-conformance regarding the use of state space results. Various linear combinations of a system's
state variables can be used to span its state space and different reconstruction methods can yield
[11]
different solutions , rendering their comparison a challenge. There should be consensus on how to
reconstruct the state space for gait dynamics in order to standardize state space methods.
6 © ISO/IEC 2021 – All rights reserved
6.4 Data flow of gait recognition
Figure 2 illustrates the components and data flow between the components in a biometric gait image
sequence processing system.
Figure 2 — Components of a gait image sequence biometric system.
Comparison methods may use conventional feature-based template sets or deep convolution neural
network (DCNN) feature vectors. After the feature vectors are generated using gait signatures and
DCNN processing then the comparison is based on one of the many basic machine learning classification
algorithms e.g. Bayesian classifier or Euclidean classifier. See A.2, Deep Convolution Neural Network
(DCNN) presentations.
6.5 Body tree concept for gait
Gait imaging systems utilize 2D recordings or 3D models for human examination and for automated gait
verification and identification. Instead of using representations as isolated entities a more organized
way is to utilize the body tree structure.
For example, multimodal biometric human verification or identification may use face features, full body
features, full body gait and head movement. The results should be fused at various levels of fusion, such
as comparison score level, feature level and decision level. Submillimetre imaging should be used to
address the problem of clothing variation effect on gait matching.
Figure 3 illustrates the possibilities offered by full body images and videos, which provide a wide
selection of biometric features for various gait-related processes.
© ISO/IEC 2021 – All rights reserved 7
Key
P pitch around the side-to-side x axis
Y yaw around the vertical y axis
R roll around the front-to-back z axis
1 pose
2 appearance
3 gait
4 structure
Figure 3 — Full body features for various processes
Standard poses, element structures and data formats help the parsing of the body tree data into body
part representations and landmarks. Parsing can be achieved using methods utilizing algorithms which
process the human body as an assembly of parts. Segmentation can be used as a pre-processing step.
Both static full body and dynamic gait cues of body biometrics may be independently used for
recognition. Fusion of static and dynamic body biometrics for gait recognition can give better results
if the combination strategy is carefully balanced and the score-summation-based rule is used, for
[14]
example .
6.6 Camera image sequence requirements
The original camera image sequence is saved whenever possible without any additional cropping,
rotation or other image processing. The full body pose shall be between 60 % and 95 % of the vertical
length of the image during enrolment. The whole-body height and width shall be visible. For video
recordings, both portrait and landscape camera orientation are acceptable.
The set of photographs shall include at least one recording of the subject in a standard walking pose:
(frontal full profile, left full profile, right full profile, back full profile). Additionally, a submillimetre
wavelength recording may be included.
Gait recognition, upper body movement recognition and full body recognition can be paired to form a
multi-mode biometric process in order to improve the performance of a biometric system. If the person's
8 © ISO/IEC 2021 – All rights reserved
facial area is not visible or the number of pixels in a video surveillance or other security camera still
image is too low, then body silhouette can be used for identification or verification purposes.
Meeting the requirements set for any camera system requires measurements to be taken and analysed.
Image sequence acquisition measurements are described in Annex C.
6.7 Gait recognition recordings
6.7.1 General
A gait recognition silhouette is the image of a person represented as a solid shape of a single colour,
usually black. The edges of a silhouette match the outline of the subject. Gait recognition walk-through
video recording is recommended to improve the performance of both gait recognition and full body
photometric recognition.
6.7.2 Gait and upper body movement image encoding
There are several image encodings which shall be used instead of non-standard formats e.g. bitmaps
defined in an ad-hoc way or ambiguous formats e.g. TIFF:
a) The JPEG image sequence in Sequential baseline (in accordance with ISO/IEC 10918-1) mode of
operation and encoded in the JFIF file format (in accordance with ISO/IEC 10918-5);
b) The JPEG-2000 image sequence in Part-1 Code Stream Format (in accordance with ISO/IEC 15444-1),
lossy or lossless, and encoded in the JP2 file format (the JPEG2000 file format);
c) The PNG image sequence in Portable Network Graphics format (in accordance with ISO/IEC 15948),
lossless, and encoded according to the Portable Network Graphics (PNG) Functional specification;
d) The MPEG-4 video in AVC/H.264, in accordance with ISO/IEC 14496-10 defined format;
e) The MP-4 video in accordance with ISO/IEC 14496-14 defined format; and
Gait Recognition Landmark Points should be determined on images before compression is applied.
Landmark Points should be included in the record format if they have been accurately determined,
thereby providing the option that these parameters do not have to be re-determined when the image
is processed for body recognition tasks. The Landmark Points should be determined by computer-
automated detection mechanisms followed by human validation when necessitated by the legal
requirements. At the moment, there are no single recommendations for the gait recognition landmark
points.
6.7.3 Gait and upper body camera image resolutions
The most frequently used frame rate in digital video recording is 25 frames per second. Pixel aspect
ratio is normally 1:1. However, in several video standards the pixel is defined as non-square. For
example, a pixel aspect ratio of 0,90 is used for NTSC to display a frame size of 720 pixels x 480 pixels
(DV) or 720 pixels x 486 pixels (D1) for 720 pixels x 540 pixels displayed in 4:3 format. Most digital still
image cameras are able to record video. Submillimetre (THz) cameras and scanners have limited image
sizes in pixels due to the terahertz wavelength resolution constraints. THz frames are typically DV size.
Image orientation is generally not a problem as JPEG EXIF metadata show the camera orientation.
MPEG-4 AVC/H.264 (ISO/IEC 14496-10) implementations for video coding allow frame extraction for
biometric sample comparison processing to take place. MPEG-4 Part 14 or MP4 is a digital multimedia
format most commonly used to store video and audio. MPEG-4 Part 14 (formally ISO/IEC 14496-14) is a
standard specified as a part of MPEG-4, in accordance with ISO/IEC 14496-1 and ISO/IEC 14496-2. MP4
is the related file format.
Table 2 shows the most common digital video formats and respective resolution, aspect ratio and pixel
size information. Figure 4 shows the VGA, HD and 4K frames placed on a single 8K frame.
© ISO/IEC 2021 – All rights reserved 9
Table 2 — Comparison of digital video recording formats
Video format name Resolution (pixel) Display aspect ratio pixels
VGA resolution 640 × 480 1.33:1 (4:3) 307 200
HD 720 p 1280 × 720 1.78:1 (16:9) 921 600
HD 1080 p 1920 × 1080 1.78:1 (16:9) 2 073 600
DCI 2K 2048 × 1080 1.90:1 (19:10) 2 211 840
UHD 4K (UHD-1) 3840 × 2160 1,78:1 (16:9) 8 294 400
DCI 4K 4096 × 2160 1,90:1 (19:10) 8 847 360
UHD 8K (UHD-2) 7680 × 4320 1,78:1 (16:9) 33 177 600
Figure 4 — Comparison of digital video frame sizes
6.8 Gait modality
6.8.1 General
Gait recognition and full body recognition can be paired to form a multi-mode biometric process in
order to improve the performance of a biometric system. If the person's facial area is not visible or
the number of pixels in a video surveillance or other security camera still image is too low, then body
silhouette can be used for identification or verification purposes.
High resolution level-50 and 4K still images are suitable even for facial biometric recognition processes.
In automated processing, it can be necessary for the parsing of the body tree to occur first, which in
practice involves the detection of all body parts and forming a body tree model. In this case the facial
and upper torso is used to get the facial image. Some face recognition programs include this feature.
For example, in order to compare two video frames, including one reference video stream of a person
seen in various full body poses and one CCTV video probe showing only the upper torso and head, it
would be the right decision to apply head movement analysis on both video streams.
6.8.2 Gait silhouette
A gait recognition silhouette is the image of a person represented as a solid shape of a single colour,
usually black. The edges of a silhouette match the outline of the subject. Gait recognition walk-through
10 © ISO/IEC 2021 – All rights reserved
video recording is recommended to improve the performance of both gait recognition and full body
photometric recognition.
To ensure that the gait sequence captures the body movement in detail, it is recommended that the
sequence be captured at the rate of 30 frames per second. This is a typical frame rate used in gait
[15] [16]
research databases such as the CMU MoBo database and USF HumanID gait database .
In order to capture all the details of a gait signature, a minimum of one full gait cycle, i.e. two full steps,
shall be captured. Figure 5 illustrates the silhouettes of the phases of one full gait cycle showing from
left to right stance phases (1, 3 and 5) and between those, swing phases (2 and 4).
Key
1 stance phase
2 swing stage
3 stance phase
4 swing stage
5 stance phase
Figure 5 — Illustration of the phases of a full gait cycle
Various automated methods have been developed for biometric gait recognition. Some methods use
the image data as an input while others only use silhouettes. Also, automated methods can use either
aligned images or non-aligned images. The capture process should allow for any method to be used for
automated recognition, therefore the gait sequence should be captured with a stationary camera.
[17]
The side view is the most discriminative view of a gait sequence. The subject should be captured
at least using side-view. The subject should be instructed to walk on a straight line perpendicular to
the camera line of sight as illustrated in Figure 6. When a tread mill is used for walking it is easier to
maintain a stationary view of the person in the middle of the frame. As an option, make a similar video
recording showing frontal and back (dorsal) views in addition to the side (lateral) view.
Figure 6 — Top view of the camera and the person’s walking path
© ISO/IEC 2021 – All rights reserved 11
By combining appearance and motion in a spatiotemporal way it is possible to achieve better results
than using a single modality in the most difficult scenarios, where there is variation in both appearance
[12], [14]
and dynamics .
Walk-through video analysis is less time-consuming if illumination and background variations
are minimized. There are methods to reduce the effects of illumination variations and dynamic
[13]
backgrounds in video surveillance video material. Background subtraction is a challenging task,
especially in complex dynamic scenes that can contain a moving background, vegetation, rippling water,
etc. It is recommended to make studio quality recordings for enrolment so that it is not necessary to
pre-process the saved video in order to correct these problems.
6.8.3 Surveillance systems
In video surveillance systems, multiple camera operation is essential for multiple capture zones. This
may include ‘layers’ of cameras along the path a target subject is expected to take to allow for multiple
detection opportunities (e.g. as the target subject walks along a corridor towards the cameras). This is
useful not only if there are issues with frame rate and dropped frames, but also to track a target subject
if an alert is triggered, especially if there is significant detection latency.
Multiple cameras are also used for any single video surveillance capture zone that is too wide for one
camera to provide sufficient resolution of the face for the required performance levels, and to try and
compensate for target subjects that may be facing different directions when traversing a particular
camera’s field of view or depth of field. Such target subjects may be deliberately trying to avoid the
cameras or they may simply be unaware of their presence.
Gait analysis is deployed for people identification in multi-camera surveillance scenarios. View-point
independent rectification is used to calculate side view coordinates for multi-camera video frames, for
example. Low frame-rate (1-5 fps) video recordings made with still image cameras or video surveillance
systems are compatible with normalized gait silhouette sequences.
Typical recording time for enrolment is in the order of 6 seconds. Ide
...
Questions, Comments and Discussion
Ask us and Technical Secretary will try to provide an answer. You can facilitate discussion about the standard in here.
Loading comments...