Voice Based Autonomous Access Control Terminals

1
Voice Based Autonomous Access Control Terminals

• HEXIUM Technical Development Co., Ltd.
• 2003

2
Table of Contents

• Introducing Hexium Ltd.
• The problem
• Access control
• Biometrics
• Voice-print properties
• Innovation
• Preliminary market analysis
• Summary

3
HEXIUM Technical Development Co., Ltd.

• Research, development, production (hardware,
  software)
• Security control, image processing, industrial
  measurement, signal processing and automation

4
The problem

1. Physical access/time control
2. Inexpensive, reliable, fast, user-friendly

5
Issues with ownership based access control

• What to identify?
• a card?
• a keyword?
• a PIN code?
• and NOT the PERSON!
• Gives rise to abuse..

6
Biometrics

• Identification of human subjects based on
• Physical parameters
• Behavioral parameters
• Examples
• Fingerprint, iris, face, ear, hand
• Signature, voice

7
Comparision of biometrics
Method Recognition User acceptance User friendliness Difficulty/method of abuse
Fingerprint Sensor dependent
Signature Learning
Face Masks
Retina/iris ---- ---- -----
Hand Sensor dependent
Voice Sensor and system dependent
8
But..Issues with biometrics

• Expensive sensors
• Poor acceptance by users due to..
• Uncomfortable use
• Mistrust
• Precision, reliability
• Recall
• Slow processing

9
Properties of voice

• Identitity verification Speaker dependent
  features (speaker verificaton) Message
  verification Message dependent features
  (speech recognition)

10
Voice-print properties
The figures show the spectogram of a sentence
spoken twice by the same person
Both the rhythm (a behavioral property) and the
spectral distribution of identical sounds
(physical parameters) stay remarkably constant.
Speech of different people show vastly different
rythmic and spectral patterns.
People are generally good at copying behavioral
patterns (a behavioral property), but is is
impossible for them to copy the physical layout
of sounds.
11
Issues with speaker verification

• Noisy environment
• People staying close
• Takes time (dialogue)
• Goal scaleable, multi-layer system
• FAR 0.01-0.0001
• FRR small (below 1)

FAR false acceptance rate FRR false reject rate
12
Innovation

• Advanced noise filtering
• Liveness verification using array microphones
• Optimized implementations
• Robust dialogue design minimizing the error rate
• Multi-model user interface
• Open system architecture

13
Preliminary market analysis
We are here..

• IDC
• Access control, 2002 USD 4.2M
• Biometrics, only banks, 2004 USD 1,800M

14
Summary

• Advanced, autonomous voice based access control
  terminals
• Competitive technology
• Better acceptance than that of competing
  biometrics technologies
• Growing market demand

15
Contact

• HEXIUM Technical Development Co., Ltd.
• H-1134 Budapest, Váci út 51/b, HUNGARY
• Phone [phone]
• Fax [phone]
• E-mail mail_at_hexium.hu
  tamas.nagy_at_hexium.hu
  csaba.szepesvari_at_hexium.hu
• Honlap www.hexium.hu www.hexium.hu/voicetermi
  nals