Hybrid Secure-Entry System

1
Hybrid Secure-Entry System

• By Group 8
• Mark Bronsberg Nick Stout
• ECE 445 Senior Design Project December 1st, 2005

2
Introduction

• Security is a major issue in the 21st century
• e.g., car alarms, power door locks, card readers,
  even video surveillance
• Biometrics are a good key
• Hard to fake, nothing to carry
• Radio Frequency IDentification (RFID) is
  currently a very popular technology
• I-Pass, Inventory Management, Theft Prevention

3
Objective
We wanted to use a combination of portable
authentication and biometric verification to
form a more secure, stable, and easy-to-use
security system. In particular, we planned on
using RFID tags and a users voice to confirm a
valid entry.
4
System Description
Antenna
5
System Description

• Hardware interfacing
• TI RFID Equipment
• PC
• Relay Driver
• LCD
• Software interfacing
• MATLAB

Image acquired from Texas Instruments
(http//www.ti.com/rfid/graphics/productImages/stu
-650a.jpg)
6
Features

• Uses RFID Authentication
• RFID Tags are passive - no need to replace
  batteries
• RFID Tags have a unique ID for users to be
  recognized, and cannot easily be spoofed
• Hands-free larger tags can be easily read up to
  1.5 feet away from antenna through pockets.

7
Features

• Uses Biometric (Voice/Face) Verification
• Neither voice nor face can be easily faked
• Verification is fast Less than 2 seconds
• Requires little work by the user
• Items cannot be forgotten at home

8
Features

• Other Security Features
• Times out during verification (if no data is
  received)
• Too many attempts ? Locks user out
• Easily adaptable for more forms of security with
  serial interfacing
• Name/password database is stored on PIC for
  ultimate security

9
General Flow of Process
Start
RFID ready
Ping RFID reader
Set lockout bit 1
Valid RFID?
No
Yes
Greet user, Request PW
Check Password
Timeout
Lockout
Unlock relay
Wait 8s
Lock relay
10
Hardware Overview
11
Hardware Overview
LCD
Relay Driver
PIC16F
RS232 Line Driver
10 MHz Oscillator
12
Hardware Overview
13
Hardware Overview

• PIC communicates over two UARTs
• Hardware UART communicates with RFID Reader
• Software UART communicates to PC
• MAX232 driver raises 5V TTL levels to 10V RS232
  levels
• PIC sends user instruction to LCD buffer
• PIC opens/closes relay contacts (with relay
  driver circuit)

14
Relay Driver

• PIC sends a high (5V) through resistor
• Current is amplified by BJT
• Current flows through coil and closes relay
  contacts
• Diode stops flyback current from destroying other
  components

15
RFID Reader

• Reader runs in Buffered Read mode
• Operates at 13.56 MHz
• Power usage varies by activity
• Scanning peaks around 4.3W, powering up peaks
  around 24W
• Reader is responsible for activating the
  transponders and receiving their data.
• Requires matched-Z antenna (50 ?)

16
RFID Tags

• Tags operate at 13.56 MHz
• Activated by Reader
• Receives power from antenna and transmits back
• Comes in 4 sizes for different configurations
• Tags can store packets of data

Image acquired from Texas Instruments
(http//www.ti.com/rfid/docs/manuals/refmanuals/hf
i_inlays_ref_guide.pdf)
17
RFID Antenna

• 50 ? characteristic impedence (matched)
• 13.56 MHz operating frequency
• Very placement-sensitive
• Near-field
• ? 22 m
• Tested read distance 0.46 m

18
Tag Range
Tags held near antenna
Tags kept in pocket

• Antenna can receive larger tags up to about 1.5
  feet
• Smaller tags shouldnt be used in security
  system, as they dont meet our demands (1
  minimal read distance)

19
Biometric Authentication

• Originally planned to use the TI-54x DSP for
  voice recognition
• embedded, portable system
• 1/8 audio connection
• serial port
• Misplaced emphasis on learning ECE 420 labs and
  assembly lost time
• Chose to implement design with MATLAB and add
  face recognition

20
Cepstral Analysis
Cepstral Analysis for Voice Recognition
speech
FFT
log10
Mel
FFT -1
cepstrum
21
Cepstral Analysis
Cepstral Analysis for Voice Recognition
speech
FFT
log10
Mel
FFT -1
cepstrum

• Uses source-filter model of speech production

• s speech signal heard by ear/system
• h transfer function of vocal tract, glottis
• e excitation signal (pitch of speech)

22
Cepstral Analysis
Cepstral Analysis for Voice Recognition
speech
FFT
log10
Mel
FFT -1
cepstrum

• Uses source-filter model of speech production

23
Cepstral Analysis
Cepstral Analysis for Voice Recognition
speech
FFT
log10
Mel
FFT -1
cepstrum

• Uses source-filter model of speech production

24
Cepstral Analysis
Cepstral Analysis for Voice Recognition
speech
FFT
log10
Mel
FFT -1
cepstrum

• Convert to Mel-frequency scale
• approximates the manner in which humans hear
  sound
• Low frequencies variations are distinct
• High frequencies variations are less clear

25
Cepstral Analysis
26
Cepstral Analysis
Cepstral Analysis for Voice Recognition
speech
FFT
log10
Mel
FFT -1
cepstrum

• Uses source-filter model of speech production

• It is typical to use the first 12 or 13 cepstral
  coefficients.

27
Cepstral Analysis
28
Cepstral Analysis

• cepstral coefficients ? identification vector
  (usually averaged)
• The system learns each user by storing ID
  vectors
• New recordings are compared to this ID vector by
  calculating the Euclidean distance.
• We set a threshold of ev lt 1.0

29
Cepstral Analysis

• What went wrong with our demo?
• Misunderstanding of the theory
• Cepstral coefficients not defined in most
  papers usually understood to mean cepstral
  values
• We incorrectly understood them to be the
  locations of the largest peaks in the cepstrum

30
Face Recognition

• Requires even less work by the user than voice
  recognition not affected by illness
• Methods
• Feature extraction, measurements
• 2-D Fourier analysis and correlation
• Eigenfaces

31
The Method of Eigenfaces

• M. Turk and A. Pentland, Eigenfaces for
  Recognition, Journal of Cognitive Neuroscience,
  March 1991.
• Based on principal component analysis (PCA)

32
The Method of Eigenfaces

• PCA, an easily-visualized 2-dimensional case

• We have sets X and Y

Image acquired from http//www.cs.otago.ac.nz/cos
c453/student_tutorials/principal_components.pdf
33
The Method of Eigenfaces

• PCA, an easily-visualized 2-dimensional case

• We have sets X and Y
• Find covariance matrix
• Compute eigenvectors
• Decreasing variance

Image acquired from http//www.cs.otago.ac.nz/cos
c453/student_tutorials/principal_components.pdf
34
The Method of Eigenfaces

• Forming eigenfaces for images
• Training set of 16 images G1,,G16. Average Y
• Find covariance matrix C, eigenvectors uk
• These eigenvectors are called the eigenfaces
• These orthonormal vectors form a subspace
• Project each new image, G, onto this face space
• wk ukT(G - Y) This operation is fast.
• W w1, , w16

35
The Method of Eigenfaces

• Each person has an W vector, learned by the
  system through a few initial images
• We use Euclidean distances, as we did for voice
  recognition
• We use an ef 65
• Distances between different users are typically
  gt150

36
The Method of Eigenfaces
The Average Face, Y
37
Some Eigenfaces
9
7
14
12
38
13
39
Voices vs. Faces

• Voice Recognition
• Fairly robust
• Difficult to remove background noise
• Limited to 12 or 13 cepstral coefficients
• Face Recognition
• More robust
• Easier to remove background
• Limited to M eigenfaces (M training images)

40
Successes Failures

• Stable security program on PIC
• Interfaces easily with two separate devices
• Can drive a high current device (i.e. electronic
  lock)
• Can accurately recognize user voices and faces
• Biometric verification is fast
• Not easily programmable
• Troubles with learning DSP hardware
• Interfacing between MATLAB and PIC incomplete
• Requires lab power supply

41
Challenges

• 1 Challenge Time management
• Communication with RFID Reader in Buffered Read
  Mode
• Secrets and mysteries of PIC programming
• Secrets and mysteries of DSP assembly
• Understanding of the algorithms
• Fighting wiring impatience

42
Future Developments

• Programmability by administrator(s)
• Case with power supply
• MATLAB code ported to DSP for portability
• Further testing of fully integrated system
• Voice/Face verification (matching RFID tags)
• Electronic door lock attached to relay
• More secure biometrics
• Iris or retinal scan

43
Credits

• Alex Spektor
• Professor Swenson
• Professor Douglas Jones
• Lindsay I. Smith (PCA tutorial)
• CCS Forum Users
• Ttelemah PCM Programmer
• Tony Wilson from Texas
• Instruments

44
Questions