AI in Identification of Individuals

1
AI in Identification of Individuals

• By Nick, Martin, Michael, Owen, Peter, Andre Tom

2
AI (Artificial Intelligence)

• the science and engineering of making
  intelligent machines

• Used in overlapping fields
• Research is concerned with producing machines to
  automate tasks of intelligent behaviour.

3
AI Time-line

• It has been around for years
• Has been progressing a lot faster more recently
• AI Timeline (off the BBC website)

4
Uses of AI in recognition

• Types
• Speech
• Face
• Fingerprint
• Natural language
• Handwriting
• ...and many more!

5
Uses/Advantages

• Today we're going to look at
• Face, Fingerprint Speech Recognition
• Different ways of achieving these and how they
  work
• Their practical uses / advantages and
  disadvantages

6
Facial Recognition

• By Nick Martin

7
Face recognition - Overview

• Biometrics
• Identify a person
• Humans are very good at it
• Quickly identify somebody
• Works from a distance

8
Face recognition - How it works

• Current Systems
• Recognise a face in image
• Measure up to 80 points on the face
• Distance between eyes
• Width of nose
• Depth of eye sockets
• Cheekbones
• Jaw line
• Chin

9
Face recognition - How it works

• Compares face print to database
• Can return a match based on only 20 points
• Returns best match
• Match verified by an operator

10
Face recognition - Standards

• 2D Image Standards
• Lighting
• Pose
• Facial Expression

11
Face recognition - Benefits

• Cannot be forgotten
• Works from a distance
• Can be used without peoples knowledge
• Can identify a person quickly

12
Face recognition Current uses

• Known uses
• CCTV systems
• Eliminating voter fraud
• Security ID
• Computer automation
• Biometric passports
• Crime solving

13
Face recognition Current problems

• Faces must point towards camera
• Image quality
• Very similar looking people - Twins?
• 2D only Photographs
• Image altering
• Make up
• Ageing
• Plastic Surgery

14
Face recognition 2.5D

• 2.5D
• 2D measurements
• Mapping to 3D blank model
• Semi 3D model (2.5D)
• Inconsistencies
• Facial Shape
• Feature Positions

15
Face recognition Future uses

• 3D Face printing
• Integration with other biometric systems

16
Face recognition 3D

• Video
• Similar to 2D
• Multiple images of face
• Create 3D model

17
Face recognition - Ethics

• Identifying and monitoring people without their
  knowledge
• Automatically adding to database
• Security of facial information
• Wrongful arrest/conviction

18
Face recognition - Summary

• Summary
• High computational cost
• Requires database access
• Potential inaccuracies
• Will be a part of day to day life

19
Fingerprint Recognition

• Michael Squance and Owen Jones

20
What is a Fingerprint?

• A fingerprint is an impression of the friction
  ridges of all or any part of the finger.
• Everyone's fingerprints are different!
• Why do we use them?
• Crime Scene Analysis
• Security Systems
• Register Systems

21
How is this achieved?

• There are two main technologies used in
  fingerprint recognition
• Pattern/ image based
• Image on image comparison

22

• Minutia-based
• Comparison of distinguishing features of a
  fingerprint

But how do we get these into a computer?
23
Methods of data capture

• There are three different methods used in this
  process
• Optical
• Ultrasonic
• Capacitance

24
Is this effective?

• Fingerprint recognition has been used for 100
  years
• Only a few recorded cases of it being unreliable
• E.g. Stephan Cowans
• Claimed to be 10x more effective than other
  methods e.g. DNA
• We believe its very effective method of
  identification

25
Speech Recognition

• By Peter Andre

26
Speech Recognition

• What Is Speech Recognition?
• How Do They Work?
• How Good Are Current Systems?
• What They Do Well and Badly?
• Who Uses Them?

27
What Is Speech Recognition

• Mapping of an Acoustic Voice Signal
• Conversion Of Speech to Text
• Understanding Speech to Form an Abstract Meaning.
• Person Recognition.

28
Vocabulary Size

• Small vocabulary - tens of words
• Medium vocabulary - hundreds of words
• Large vocabulary - thousands of words
• Very-large vocabulary - tens of thousands of
  words.

29
Dependent, Adaptive or Independent?

• A speaker dependent system is-
• A speaker independent system is-
• A speaker adaptive system is-

30
Isolated Word or Continuous Speech?

• Isolated Word
• Continuous Speech

31
Recap

• Looked at different ways of using AI in
  recognition.
• Various uses.
• Advantages and limitations (quickly becoming
  less)
• Maybe combine systems?
• http//users.aber.ac.uk/tjs6/cs18110/grouppresenta
  tion.shtml