An Introduction to

S-Transform for
Time-Frequency Analysis

S.K. Steve Chang

SKC-2009
Question: Do we live in…

Time-Domain
or
Frequency-Domain?

SKC-2009
 Time Domain vs
Frequency Domain
Waveform

0.1 0.9
Time (Seconds)

Spectrum
Time
information
is lost.

2000 4000 6000

8000
Frequency (HZ)

6000

4000

2000

0.1 0.3 0.5 0.7 0.9

Time (Seconds)

SKC-2009
 Fourier Transform &
Short Time Fourier Transform
FourierTransform{x(t)} 


 i2 ft
X( f )  x(t)e dt


STFT{x(t)} 


 i2 ft
X( , f )  
x(t)W (t   )e dt

Window
SKC-2009
 STFT Fourier Transform

Spectrogram{x(t)}  X( , f )
2

FourierTransform{x(t)} 



X( f )  X( , f )d


SKC-2009
STFT: Fixed Window, Fixed Resolution

Window Length

Frequency Resolution

Time Resolution

SKC-2009
Example (400 msec sample)

SKC-2009
(T=25 ms) Poor Frequency Resolution

SKC-2009
(T=125 ms) Fixed Resolution

SKC-2009
(T=375 ms) Fixed Resolution

SKC-2009
 (T=1000 ms) Poor High
Frequency Time Resolution

SKC-2009
 S-Transform &
Short Time Fourier Transform
S  Transform{x(t)} 
(t )2 f 2
f 

 i2 ft
S( , f )  x(t)e 2
e dt
2 

STFT{x(t)} 


 i2 ft
X( , f )  
x(t)W (t   )e dt

SKC-2009
Gaussian Window for S-Transform


 i2ft
S(, f )  
x(t)W (t   , f )e dt

t2 f 2
f 
W (t, f )  e 2
2

Function of Frequency
SKC-2009
Gaussian Window for S-Transform

High Frequency
Time Shifted

Low Frequency

SKC-2009
 X( f ) and x(t) from S(, f )



X( f )  S( , f )d
 


  
  i2 ft
x(t)   
S( , f )d e df

SKC-2009
Two Chirps and Two Bursts
S-T

STFT Wigner

SKC-2009
S-Transform and Wavelet Transform
Continuous  Wavelet  Transform{x(t)}
t 
 C( ,d)   x(t)M(

)dt
d


 C( , f )  x(t)M((t   ) f )dt

t2 f 2
f 
i2 ft
Mother Wavelet: M(t, f )  e 2
e
2
i2 f
S( , f )  e C( , f )
SKC-2009
S-Transform is not Wavelet Transform
Mother Wavelet does not have zero mean:





M(t, f )dt  0
Morlet’s Wavelet  M(t, f )  K( f )
i2 f
S( , f )  e C( , f )
 Phase Correction
SKC-2009
 Generalized S-Transform


 i2ft
S(, f , p)  
x(t)W (t   , f , p)e dt
t2 f 2
f 
W (t, f , p)  e 2p2

2 p
Standard Deviation Resolution
t p 
1 p p Time Frequency
2 f
1 f
f p 
2 p
SKC-2009
Generalized S-Transform: Bi-Gaussian


 i2ft
S(, f , p)  
x(t)W (t   , f , p)e dt
t2 f 2
f 
2
W (t, f , p)  e 2 p(t )2

2 (P1  P2 )
where
 P1, t  0
p(t)
 P2 , t  0

SKC-2009
Generalized S-Transform: Bi-Gaussian

Sharper
Front Edge
For Arrival
Time

Sacrifice
Frequency
Resolution

SKC-2009
Generalized S-Transform: Bi-Gaussian
Sharper Arrival Time

SKC-2009
Generalized S-Transform: Your Own


 i2ft
S(, f , p)  
x(t)W (t   , f , p)e dt
f 2 p(t )2
f 2 
W (t, f , p)  e 2
2 (P1  P2 )
where
P1  P2 P1  P2
p(t)  (t  k)  (t  k) 2  P32
2P1P2 2P1P2
(P1  P2 ) 2 P32
k
4P1P2
SKC-2009
Generalized S-Transform: Your Own


 i2ft
S(, f , p)  
x(t)W (t   , f , p)e dt

Constraint:



W (t, f , p)dt  1 Or




W (t   , f , p)dt  1 for all 


SKC-2009
 Examples - Applications
Engine Knock Pressure STFT

S-Transform with New Window S-Transform

SKC-2009
Power Disturbance Detection
Power Disturbance Power Disturbance

Contour Contour

Stationary Phase Stationary Phase

SKC-2009
 Earthquake Detection
Noisy Signal Time-Frequency Filter
Inverse S-Transform

Earthquake Signal Earthquake Signal

SKC-2009
 Earthquake Seismology
Original Data

Filter 1

Filter 2

S-Filters SKC-2009
 Seismic Data
Seismic
Trace S- STFT
Transform

SKC-2009
Medical: Forearm Blood Flow
Rest

Activities
Occlusion

SKC-2009
 Geomagnetic Storm
Induced Current

SKC-2009
Gear Vibration Signal Decoposition

Helicopter Rotor Transient

SKC-2009
 Image Analysis
Original Image Dominant Frequency

SKC-2009
 S-Transform
We live in Time-Frequency World.

• Similar to Short-Time Fourier Transform

SKC-2009