Skip to content

Introduction

Biometrics is the process of recognising people by personal characteristics, typically through computer vision. This module gives an overview of different techniques used in biometrics, including those from computer vision and image processing.

We also cover performance factors in these systems: how well they work, the limitations, and the advantages of these systems. The course also covers where current research is and how these systems are being improved.

Biometrics is a good interdisciplinary field, taking knowledge from programming, mathematics (discrete, continuous, and statistics), and different types of hardware. Biometrics can be used to recognise people based on unique and permanent personal identifiers.

Lectures

Lectures appear in chronological order, both here and in the sidebar. For the coursework, you will need the lectures on gait, and the computer vision lectures, most likely.

Reading

The course covers the main points and aims to expose us to lots of different systems. Further reading is guided through:

  • Citations in the lectures,
  • Feature Extraction & Image Processing for Computer Vision (PDF, Archived),
  • Famous techniques, and
  • Personal choice
  • Further textbooks:
    • Jain, Biometrics Personal Identification in Networked Society
    • Jain, Handbook of Biometrics
    • Li, Encyclopedia of Biometrics

Broad Definition

A biometric is a unique, personal identifier. This is something that we have, not something that we know (like a password), or something that we carry (like a key or U2F device). When selecting features for biometrics, we need to consider four things:

  • Universality -- do we all have that trait?
  • Acceptability -- are we prepared to reveal it, or happy for it to be measured?
  • Uniqueness -- is it unique to an individual?
  • Repeatability -- are the measurements the same at different times?

The process of enrolment and verification is broadly similar to a machine learning classifier, in that we take the data, do some processing, and extract features. Unlike a machine learning model, which updates weights and biases through gradient descent, we store the feature to a database.

graph LR

subgraph Enrolment
direction LR
A(Data) --> B(Processing) --> C(Features)
end
C --> D[(Database)];
D --> H;
subgraph Verification
E(Data) --> F(Processing) --> G(Features) --> H(Matching);
H -->|Verification| I[Accept];
H -->|Recognition| J[Identity];
end

When going through the verification (or recognition, identification) process, we extract the same features from the challenge data, then compare to database values. Verification is the process of accepting a person, whereas recognition is matching that person to a specific identity.

Computer Vision

Human vision is similar in concept to computer vision, in that a human eye is a form of sensor, and the brain processes and stores the information seen. It works differently to a camera, in that the eye is not comprised of pixels, and processing is carried out by neurons, as opposed to transistors and logic gates.

Computer vision is good for non-contact biometrics. It can be used where data can be extracted in an image-like fashion. A camera works, as does a capacitative sensor for fingerprint recognition. We can look at an image in either a model-based way, extracting the features and points from the image, or use image based measures, such as taking transforms or mappings of different images. Fingerprint scanners, palm scanners, and iris detection are all good examples of computer vision techniques.

These techniques are then used at immigration checkpoints for the fast passage of a specific person, or can be used at places such as Disney to verify that a person using a wristband is the same.

Administrivia

The course is split into nine sections:

  1. Introduction to biometrics. Applications of biometrics. Overview of computer vision methods.
  2. Computer vision and image processing. Automated analysis of computer images. From images to measurements.
  3. Face and fingerprint biometrics. Holistic and model-based approaches. Applications of vision-based recognition.
  4. Gait biometrics. Recognition by walking and running. Moving object recognition and description.
  5. Identification through the ages: history of biometrics and (forensic) identification.
  6. Iris recognition. Iris image acquisition and processing. Performance limits and evaluation.
  7. Demonstration. How do biometrics systems really work? Can we recognize people?
  8. New modalities and current research. Performance limits, how will they be resolved. Other info.
  9. Palm Biometrics Oldest biometric and how does it work.

Coursework

The coursework accounts for 30% of the total module grade and is on Gait biometrics. The coursework is due on the 14th May 2024. The coursework specification can be found here for those at the University, and requires a 2000 word report and code samples. Main page here.

Lectures

Info

The lectures are not linear, and jump around throughout each week. Topics 5,7, and 8 also do not appear to have slides available. This section is mainly for me to see what slides were covered on what days, so I can keep up with what to look at. Panopto links are for internal use only for this cohort and will otherwise not work.

By Date

Date Topic Lec No. Slide Nos. Panopto Done?
Some Missing
2024-02-20T13:00 Computer Vision 2 43-56 (13) Link
2024-02-22T13:00 Face & Fingerprints 3 30 (1) Link ✓ (Not brilliant notes)
2024-02-23T16:00 Face & Fingerprints 3 30-39 (9) Link ✓ (Not brilliant notes)
2024-02-27T13:00 Computer Vision 2 56-74 (18) Link
2024-02-29T13:00 Face & Fingerprints 3 39-51 (22) Link
2024-03-01T16:00 Face & Fingerprints 3 51-59 (8) Link to 27 mins
Gait Biometrics 4 1-12 (12) Link 27 mins onwards
2024-03-05T13:00 Computer Vision 2 75-87 (12) Link
2024-03-07T13:00 Computer Vision 2 87-103 (16) Link ✓ (Missing a bit of a derivation)
2024-03-08T16:00 Computer Vision 2 104-115 (11) Link
2024-03-12T13:00 Gait Biometrics 4 13-35 (22) Link
2024-03-14T13:00 Gait Biometrics 4 35-39 (4) Link
2024-03-15T16:00 Computer Vision 2 116-128 (12) Link ✓ (Not brilliant notes)
2024-03-19T13:00 Coursework Introduction -- -- Link to 14 mins ✓ (Notes on spec, not CW itself)
Computer Vision 2 117-130 (13) Link 14 mins onwards ✓ (Summary of last comp vision)
2024-03-21T13:00 Gait Biometrics 4 38-54 (16) Link
2024-03-22T13:00 Gait Biometrics 4 54-59 (5) Link
Easter Break -- -- -- -- --
2024-04-23T13:00 Iris Biometrics 6 1-26 (26) Link
2024-04-25T11:00 Iris Biometrics 6 26-29 (3)

By Topic

Date Topic Lec No. Slide Nos. Panopto Done?
Some Missing
2024-02-20T13:00 Computer Vision 2 43-56 (13) Link
2024-02-27T13:00 Computer Vision 2 56-74 (18) Link
2024-03-05T13:00 Computer Vision 2 75-87 (12) Link
2024-03-07T13:00 Computer Vision 2 87-103 (16) Link ✓ (Missing a bit of a derivation)
2024-03-08T16:00 Computer Vision 2 104-115 (11) Link
2024-03-15T16:00 Computer Vision 2 116-128 (12) Link ✓ (Not brilliant notes)
2024-03-19T13:00 Computer Vision 2 117-130 (13) Link 14 mins onwards ✓ (Summary of last comp vision)
2024-02-22T13:00 Face & Fingerprints 3 30 (1) Link ✓ (Not brilliant notes)
2024-02-23T16:00 Face & Fingerprints 3 30-39 (9) Link ✓ (Not brilliant notes)
2024-02-29T13:00 Face & Fingerprints 3 39-51 (22) Link
2024-03-01T16:00 Face & Fingerprints 3 51-59 (8) Link to 27 mins
2024-03-01T16:00 Gait Biometrics 4 1-12 (12) Link 27 mins onwards
2024-03-12T13:00 Gait Biometrics 4 13-35 (22) Link
2024-03-14T13:00 Gait Biometrics 4 35-39 (4) Link
2024-03-21T13:00 Gait Biometrics 4 38-54 (16) Link
2024-03-22T13:00 Gait Biometrics 4 54-59 (5) Link
2024-04-23T13:00 Iris Biometrics 6 1-26 (26) Link
2024-04-25T11:00 Iris Biometrics 6 26-29 (3)
2024-03-19T13:00 Coursework Introduction -- -- Link to 14 mins ✓ (Notes on spec, not CW itself)

Last updated: 2024-04-20

Comments