Real Time Digital Watermarking System for Audio Signals

1
Real Time Digital Watermarking System for Audio
Signals
Technion - Israel Institute of Technology
Department of Electrical Engineering Signal and
Image Processing Laboratory

• Yuval Cassuto and Michael Lustig
• Supervisor Shay Mizrachi

2
Motivation - Music Piracy
World Recording industry US 37 Billion
Internet music is virtually a 100 pirate medium
According to IFPI report
3
The Watermarking Concept
Signature Embedding
owner signature 1345234
4
Signature Detection
Adversary
owner signature 1345234
5
Signature Requirements

• Embedded
• Inaudible
• Public Algorithm
• Damaged Signature?Damaged Audio
• Resolve Deadlock ? Keep Original

6
The watermarking problem

• Generation of a unique, robust and hidden
  signature
• Find an appropriate embedding method and location
  
• Embedding

7
Signature Embedding Algroithm
Output
Output
OriginalSignalSegment
W
Watermark Coloring
Frequency Masking(Psychoacoustic model)
Watermark Coloring
FFT
W
Pseudo RandomNoise Generation
Local KeyCalculation
Local KeyCalculation
Pseudo RandomNoise Generation
1345234
Owners Key
8
Frequency Masking
Signal Spectrum
9
Watermark Coloring
Masking Threshold
White Watermark
Colored Watermark
10
Signature Detection Algorithm
W
Correlation Threshold
Gain Matching
Tested
Decision
Signal
W
Signature Calculation
Original
Signal
1345234
Owner's
Key
11
Signature Detection Algorithm
W
Gain Matching
Correlation Threshold
W
Tested
Decision
Signal
W
Signature Calculation
Original
Signal
1345234
Owner's
Key
12
Signature Detection Algorithm
W
Correlation Threshold
Gain Matching
W
Tested
Decision
Signal
W
Signature Calculation
Original
Signal
1345234
Owner's
Key
13
Real Time Applications
Why Real-time?
14
Real-Time Implementation

• The Problem
• In the Windows application, watermark embedding
  time is x8 longer than playing time (_at_44.1KHz).
• The Solution

15
Development Phases

• MATLABTM simulation
• PC application Embedding Detection
• TMS320C54x DSP embedding implementation.
• TIGER 5410/PC real-time embedding application.

16
DSP Implementation Challenges

• Fixed Point
• Speed
• Memory
• Accuracy
• Architecture Utilization
• Parallel Execution

• Optimization
• Capacity
• I/O Synchronization

17
Algorithm Specific Implementation Challenges

• Adaptive Masking Threshold
• Log-Scale Spectrum Calculation
• Identifying Masking Components.
• Calculating Masking Curves.
• Watermark Embedding
• Creating The Watermark.
• Coloring The Watermark.

18
Adaptive Masking Threshold
115 frames/sec!
19
Masking Threshold Implementation - masking curves

• Challenge
• Masking curves are Not Linear
• Not Shift-Invariant

• Implementation
• Static Bark Addressing
• Constructing efficient Look-Up Tables

Conditional Operations ? CPU time
20
Masking Threshold Exponential Model

• Challenge
• Components and Spectrum are in Log-Scale
• Threshold filter is in Linear-Scale.
• Wide range of values.
• Vast usage of Exp?Log Transforms (more than
  200,000 per second).

21
Masking Threshold Exponent Implementation

• Calculate ex , x is within a wide range.
• First Approach
• Taylor approximation.

• Second Approach
• Look-Up table

217 words, too Big , too slow

• Hybrid Approach
• exp(ab) exp(a)exp(b)
• a ?Integer , small LUT
• b ?fraction 0,1) Taylor app.

22
Implementation ChallengesWatermark Coloring

• Challenge
• Watermark Coloring

• Implementation
• Time Domain ? Convolution O(n2)
• Frequency Domain ? Multiplication O(n)

?
23
Watermark Coloring In Frequency Domain

• Facts
• No Zero-Padding is done.
• Frequency Domain ? Cyclic Convolution

• Explanation
• Watermark is a Pseudo Random Noise
• Watermark requirements still achieved
• Ability to regenerate the same colored WM using
  the source and a public key.
• WM spectrum matches Psycho-Acoustic model
  (Inaudibility not affected) .

Filtering in frequency domain is appropriate!
24
Real-Time optimization

• Speed
• Psycho acoustic model tables dB ? Ln.
• PN26 Averaging 2x indices, Divisions ? shift.
• Optimized SQRT,EXP and LOG assembly
  implementations.
• Fixed Point
• Optimal representation to each block.
• 32 bit operations normalized and saved using
  dynamic Q 16bit representations.
• Memory
• Optimization of memory ? only fast internal
  memory used.

25
System Schematics
PC
DSP board
Data via HPI
ISA BUS
Host Applicationsoftware
DSP Application software
Sync. By interrupts and status/control registers
26
Modes of operation
Live mode Audio input Audio output
interrupt
interrupt
DAC
ADC
input
output
27
Modes of operation
Live mode Audio input Audio output
interrupt
interrupt
DAC
ADC
input
output
28
Conclusion

• Speed - 10 times faster than the PC application
• Quality - Strong inaudible watermark (-22dB)
• Completeness - Full DSP and HOST apps.
• Portability - Low power , low cost DSP (C54)
• Capacity - Easily upgradeable to a Multi-channel
  system
• Innovation - Non-Standardized field
• Commercial Value - Huge potential market

29
Acknowledgements

• The Signal and Image processing lab staff headed
  by Prof. Malah and Mr. Peleg - for the technical
  assistance
• TI - for the equipment, support , encouragement
  and invitations to ICASSP2000 and the 3rd
  European DSP Education and Research Conference .

30
(No Transcript)