Randomly Modulated Periodic Signals in Australia’s National
ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle...
Transcript of ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle...
![Page 1: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/1.jpg)
BME 310 Biomedical Computing -J.Schesser
316
Computing
Lecture #13Chapter 13
![Page 2: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/2.jpg)
BME 310 Biomedical Computing -J.Schesser
317
What Is this Course All About ?
• To Gain an Appreciation of the Various Types of Signals and Systems
• To Analyze The Various Types of Systems
• To Learn the Skills and Tools needed to Perform These Analyses.
• To Understand How Computers Process Signals and Systems
![Page 3: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/3.jpg)
BME 310 Biomedical Computing -J.Schesser
318
What did we learn so far• Learned about Signals and Systems
– Continuous-time vs Discrete-time– Sinusoids– Complex Exponentials– Periodic Signals
• How to analyze them– Sampling– Time Domain– Frequency Domain
• How to Process Them– Filters
![Page 4: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/4.jpg)
BME 310 Biomedical Computing -J.Schesser
319
What do we still have left to learn
• How does a Computer handle signals?– Sinusoids– Bio Med Signals
• The Computer can’t handle Continuous-time signals
• The Computer must first sample the signal
![Page 5: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/5.jpg)
BME 310 Biomedical Computing -J.Schesser
320
Some more Background
• We saw that the Fourier Series can be used to handle any periodic signal since it can be decomposed into frequency components.
• But most signals are not periodic– ECG, EEG, EMG, etc.– Voice Signals– Video Signals
![Page 6: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/6.jpg)
BME 310 Biomedical Computing -J.Schesser
321
Fourier Transform • We can handle non-periodic signals in a similar
fashion as periodic signals• That is, we can decompose them into frequency
components• However, the way we get there is different than the
way we use for periodic signals• Periodic signal frequency decomposition use the
Fourier Series which generates a frequency spectrum• Non-periodic signal frequency decomposition use the
Fourier Transform which generates a frequency DENSITY spectrum.
![Page 7: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/7.jpg)
BME 310 Biomedical Computing -J.Schesser
322
Fourier Analysis and Fourier Transform
• Recall this is Fourier Series
• Here’s what the Fourier Transform looks like for continuous signals, CTFT:
oo
ko
T ktT
j
o
k
o
okkj
kkook
kokok
tkfjk
dtetxT
a
TfaaeAaaAktfAAeatx
0
)2(
1
2
)(1
1*;;21; where;)2cos()(
dejXtx
dtetxjX
tj
tj
)(21)(
)()(
![Page 8: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/8.jpg)
BME 310 Biomedical Computing -J.Schesser
323
Fourier Transforms vs. Fourier Series• Note that X(jω) is a Spectral Density function; that is, if x(t) is
voltage, then X(jω) is volts/rad. – Note in Fourier series analysis, ak would also be volts is x(t) is voltage.
• Note that X(jω) is a continuous function of w and the limits of integration are over all values of t.– Note in Fourier series analysis, ak is a discrete function of kfo (and are
fo Hz apart) and the limits of summation are over one period of x(t)• A proof of how X(jω) is formulated is beyond our scope but,
briefly, X(jω) can be obtained by starting with the Fourier Series of x(t) (as if it were periodic) and letting fo go to zero (i.e., To goes to infinity which make the second repetition of x(t) move to infinity and makes x(t) non-periodic. – This would make the spectral components move closer to each other
(infinitely closer – make 2πkfo a continuous variable ω)– This will also make ak approach zero but the ratio of ak/fo, which is a
spectral density, remains finite.
![Page 9: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/9.jpg)
BME 310 Biomedical Computing -J.Schesser
324
Discrete-time Fourier Transform• If this is the continuous-time Fourier Transform
• Then replacing t with nTs and the integral with a summation, then the Discrete-Time Fourier Transform, DTFT, can be shown to be :
ˆ ˆ
( ) ( )
( ) [ ]
ˆ Recalling that
sj nTs
n
j j n
n
s
X j x nT e
X e x n e
T
dtetxjX tj )()(
![Page 10: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/10.jpg)
BME 310 Biomedical Computing -J.Schesser
325
Discrete-time Fourier Transform• Note that this looks very similar to the Frequency
Response of a system
• As a matter of fact, the Fourier Transform of the Impulse response is the Frequency Response
ˆ ˆ
ˆ ˆ
0
( ) [ ]
( )
j j n
nM
j j kk
k
X e x n e
H e b e
)(][][)( ˆ
0
ˆˆˆ jM
k
kj
k
kjj eHekhekheX
![Page 11: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/11.jpg)
Discrete Fourier Transform
BME 310 Biomedical Computing -J.Schesser
326
The DTFT yield a spectrum which is a continuous function of
X (e j ) x[n]e jn
n
How do we get around this? Sample the spectrum.
When we sampled in the time domain, we replaced t by nTs where Ts is the distance (in time) between samples.Therefore to sample in the frequency domain we replace 2 f by 2kf where f is the distance (in frequency) between spectrum samples.
Note that since Ts fs
; fs
2kf
fs
X (ej
2kffs ) X [k] x[n]e
j2kf
fsn
n
Let us assume that there are only L samples for time domain and N samples for the spectrum.
X [k] x[n]e j
2kffs
n
n0
L1
Since fs is the maxium frequency in the spectrum, then f=fs
N. This is just the resolution of the displaced spectrum.
X [k] x[n]e j
2kffs
n
n0
L1
x[n]e j
2kfsfsN
n
n0
L1
x[n]e j 2
Nkn
n0
L1
This is called the Discrete Fourier Transform
![Page 12: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/12.jpg)
BME 310 Biomedical Computing -J.Schesser
327
Discrete Fourier Transform• Since the computer can only process discrete functions of finite time, we
have to define a new Fourier Transform called the Discrete Fourier Transform, DFT.– Do not confuse this with the Discrete-time Fourier Transform, DTFT.
• It is defined as
X (k) x[n]e j 2
Nkn
n0
L1
where there are the L samples of x[n], we evaluate the Spectrum over N frequencies, i.e., 0 k N 1,
and each frequency is f apart and chose f=fs
N
since fs is the maximum frequency of the spectrum.
Therefore, f=fs
N
1NTs
. We call this the resolution of the spectrum.
![Page 13: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/13.jpg)
BME 310 Biomedical Computing -J.Schesser
328
Discrete Fourier Transform
Let's start with the DTFT: X (e j ) x[n]e jn
; Ts
Let's divide the spectrum is into N frequencies equally spaced f Hz apart (i.e., we are sampling the spectrum).
Therefore, let's define the kth sample in the frequency domain as k 2 fk 2kfwhere k goes from 1 to N .
When k N , the highest frequency in the spectrum is N 2N
To
=2Nf=2 fs .
2πfΔ 2(2πfΔ) 3(2πfΔ) N(2πfΔ)=2πfs
. . .ω
![Page 14: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/14.jpg)
BME 310 Biomedical Computing -J.Schesser
329
Discrete Fourier Transform
If fs meets the Nyquist rate, then the one-sided spectrum of x[n] = X (e j ) must end at or below fs
2.
Therefore, k kTs 2kfTs 2kfs
NTs
2kN
.
Let's substitute k for in the DTFT: X (e jk ) x[n]e jkn
n
x[n]e j 2k
Nn
n
This sum will only be a function of k. In addition, let's assume that there are L samples of x[n].
Then, we have the Discrete Fourier Transform, DFT as X [k] X (e jk ) x[n]e j 2k
Nn
n0
L1
2πfΔ 2(2πfΔ) 3(2πfΔ) N(2πfΔ)
. . .ω
2π fs /N
. . .
ω2(2π fs /N) 3(2π fs /N) N(2π fs /N)=2π fs
2π /N
. . .
2(2π /N) 3(2π /N) N(2π /N)=2π
![Page 15: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/15.jpg)
BME 310 Biomedical Computing -J.Schesser
330
Computer Processing
• Computers use the DFT to determine the spectrum of a signal x(t).
• There are different computer algorithms for processing the DFT– The most widely used algorithm is called the Fast
Fourier Transform: FFT
• Note that the DFT is just like a discrete Fourier Series in the Frequency Domain
1
0
22
][][ )(L
n
knN
j
k
ktT
j
k enxkXeatx o
![Page 16: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/16.jpg)
BME 310 Biomedical Computing -J.Schesser
331
How to Evaluate the DFTMethod 1: Expand n first, then k.
• What is the DFT for the x[n]={1, 1, 0, 0} assuming N=4
43
2
22
41
2
02
2111]3[
0111111]2[
2111]1[
211]0[
jj
jj
jj
j
ejeX
eeX
ejeX
eX
kj
kjkjkj
kjkjkjkj
n
knj
n
knj
e
eee
exexexex
enxenxkX
2
32
222
32
22
12
02
3
0
23
0
42
11
0011
]3[]2[]1[]0[
][][][
}2,0,2,2{][ 44 jj
eekX
![Page 17: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/17.jpg)
BME 310 Biomedical Computing -J.Schesser
332
Method 2: Expand k first, then n.• What is the DFT for the x[n]={1, 1, 0, 0} assuming N=4
4
29
323
0
3
0
34
2
320
3
0
24
2
2001
]3[]2[]1[]0[
][]3[
000)1(1]3[]2[]1[]0[
][]2[
j
jjjj
n
nj
jjjj
n
nj
ej
exexexex
enxX
exexexex
enxX
4
231
20
3
0
14
2
0000
3
0
04
2
3
0
42
200)(1
]3[]2[]1[]0[
][]1[
20011]3[]2[]1[]0[
][]0[
][][
j
jjjj
n
nj
jjjj
n
nj
n
knj
ej
exexexex
enxX
exexexex
enxX
enxkX
}2,0,2,2{][ 44 jj
eekX
![Page 18: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/18.jpg)
BME 310 Biomedical Computing -J.Schesser
333
Another Example: Method 1• What is the DFT for the x[n]={1, 1, 1, 0, 0, 0} assuming N=6
25 56 3
0 0
0 1 2 3 4 53 3 3 3 3 3
2 3 4 53 3 3 3 3
23 3
0 023 3
[ ] [ ] [ ]
[0] [1] [2] [3] [4] [5]
1 0 0 0
1
[0] 1 3
[1]
j kn j kn
n n
j k j k j k j k j k j k
j k j k j k j k j k
j k j k
j j
X k x n e x n e
x e x e x e x e x e x e
e e e e e
e e
X e e
X
23 3 3
2 43 3
3 63 3
4 83 3
5 103 3 3
1 1 .5 0.86 .5 0.86 1 2(0.86) 2
[2] 1 1 .5 0.86 .5 0.86 0
[3] 1 1 1 1 1
[4] 1 0
[5] 1 1 .5 0.86 .5 0.86 2
j j j
j j
j j
j j
j j j
e e j j j e
X e e j j
X e e
X e e
X e e j j e
![Page 19: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/19.jpg)
BME 310 Biomedical Computing -J.Schesser
334
Method 2• What is the DFT for the x[n]={1, 1, 1, 0, 0, 0} assuming N=6
2 25 2 1 26 3 3 3 3 3
0 020 0
3 3
2 21 13 3 3 3 3
2 2 42 23 3 3 3
23 33 3
[ ] [ ] 1 1
[0] 1 3
[1] 1 1 2
[2] 1 1 0
[3] 1 1
j kn j kn j k j k j k j k
n n
j j
j j j j j
j j j j
j j
X k x n e e e e e e
X e e
X e e e e e
X e e e e
X e e e
2
2 4 84 43 3 3 3
2 5 105 53 3 3 3 3
1
[4] 1 1 0
[5] 1 1 2
j j
j j j j
j j j j j
e
X e e e e
X e e e e e
![Page 20: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/20.jpg)
BME 310 Biomedical Computing -J.Schesser
335
What is the right hand side of the FFT.vi spectrum?
• Note that N and L are usually taken to be the same in order to easily calculate X[k].
• Note that due to this fact X[N - k]=X[-k]=X*[k]
• Therefore, X[-k] will show up as X[N – k]
][*][
][][
][][][
][][
21
0
21
0
2
21
0
21
0
)(2
1
0
2
kXkX
enxeenx
eenxenxkNX
enxkX
knN
jN
n
knN
jN
n
nj
knN
jN
n
NnN
jN
n
nkNN
j
N
n
knN
j
![Page 21: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/21.jpg)
BME 310 Biomedical Computing -J.Schesser
336
Take a look at our Example• What is the DFT for the x[n]={1, 1, 0, 0} assuming N=4
]3[]14[200)(1
]3[]2[]1[]0[
][]1[
][][
4
231
20
3
0
14
2
3
0
42
XXej
exexexex
enxX
enxkX
j
jjjj
n
nj
n
knj
![Page 22: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/22.jpg)
BME 310 Biomedical Computing -J.Schesser
337
Another Example
ˆ( )1
ˆHere we have a signal whose digitized frequency is
[ ] for 0, 1, 2, , 1
We now want to obtain the spectrum of our signal usingthe DFT. If the DFT is done correctly, we would ex
o
o
j nx n e n N
21 1ˆ( ) (2 )
0 01
ˆ[(2 ) ]
0
pect a ˆthe spectrum to show a single component at
[ ] [ ] o
o
o
N Nj kn j n j N knN
n nN
j k N nj
n
X k x n e e e
e e
![Page 23: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/23.jpg)
BME 310 Biomedical Computing -J.Schesser
338
Another Example Continued
Using the partial sum of a geometric series: k 1 L
1k0
L1
X [k] e j (1 e j[(2k N )o ]N
1 e j[(2k N )o ] ) e j (e j[(2k N )o ]N /2 (e j[(2k N )o ]N /2 e j[(2k N )o ]N /2 )e j[(2k N )o ]/2 (e j[(2k N )o ]/2 e j[(2k N )o ]/2 )
)
e j (e j[(2k N )o ]( N1)/2 sin([(2k N ) o ]N / 2)
sin([(2k N ) o ] / 2))
e je j[(2k N )o ]( N1)/2 sin([(2k N ) o ]N / 2)sin([(2k N ) o ] / 2)
e je j[(2k N )o ]( N1)/2DN (e j[(2k N )o ]N )
where
DN (e j[(2k N )o ]N ) sin([(2k N ) o ]N / 2)sin([(2k N ) o ] / 2)
, the Dirichlet function
![Page 24: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/24.jpg)
BME 310 Biomedical Computing -J.Schesser
339
Another Example Continuedˆ[(2 ) ]( 1)/2
0 0
0
ˆsin([(2 ) ] / 2)[ ]ˆsin([(2 ) ] / 2)
2ˆ ˆ2
Case 1: is NOT an integer; ˆthat is, is NOT a multiple of 2 and therefore is NOT a mu
oj k N Nj o
o
o oo o o s s s
o
o
k N NX k e ek N
k kk N f f f fN N
kN f
[(2 ) 2 ]( 1)/2
[2 ( )](
ltiple of , which is the resolution of the sampled spectrum obtained using DFT.
0 for any value of sin([(2 ) 2 ] / 2)sin([(2 ) 2 ] / 2)
o
o
s
o
j k N k N Nj o
o
j N k k Nj
f
k k kk N k N Ne ek N k N
e e
1)/2
[2 ( )]( 1)/2
sin([2 ( )] / 2)sin([2 ( )] / 2)sin[ ( )] ; for all
sin[ ( )]o
o
o
j N k k Nj o
o
N k k NN k k
k ke e kN k k
![Page 25: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/25.jpg)
BME 310 Biomedical Computing -J.Schesser
340
-20
0
20
40
-1 -0.8 -0.5 -0.3 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3
Two Cases for our Example2 (2.5)ˆsay let 2.5 and 40; 2.5 0.125
40 20 8ˆIf we do this, the figure below shows that do not have a single component at
o o
o
k N
2.5; 40
sin([ ( 2.5)])[ ]sin([ 40( 2.5)])
sin([ (0 2.5)])[0] 5.1sin([ 40(0 2.5)])
sin([ (1 2.5)]) sin([ ( 1.5)])[1] 8.5sin([ 40(1 2.5)]) sin([ 40( 1.5)])
sin([ (2 2.5)]) s[2]sin([ 40(2 2.5)])
ok N
kX kk
X
X
X
in([ ( 0.5)]) 25.5
sin([ 40( 0.5)])
sin([ (3 2.5)]) sin([ (0.5)])[3] 25.5sin([ 40(3 2.5)]) sin([ 40(0.5)])
sin([ (4 2.5)]) sin([ (1.5)])[4] 8.5sin([ 40(4 2.5)]) sin([ 40(1.5)])
etc.
X
X
![Page 26: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/26.jpg)
BME 310 Biomedical Computing -J.Schesser
341
Another Example Continuedˆ[(2 ) ]( 1)/2
0
ˆsin([(2 ) ] / 2)[ ]ˆsin([(2 ) ] / 2)
ˆCase 2: is an integer; that is, is an integer multiple of 2 and is a multiple of .the resolution of the sampled spectr
oj k N Nj o
o
o o s
k N NX k e ek N
k N f f
[(2 ) 2 ]( 1)/2
[2 ( )]( 1)/2
um obtained from the DFT. will be a non-zero integer = except when
sin([(2 ) 2 ] / 2)sin([(2 ) 2 ] / 2)
sin([2 ( )] / 2)sin
o
o
o o
j k N k N Nj o
o
j N k k Nj o
k k l k k
k N k N Ne ek N k N
N k k Ne e
[2 ( )]( 1)/2 [2 ]( 1)/2
l 0
([2 ( )] / 20sin[ ( )] sin[ ]
sin[ ( )] sin[ ]0 since the sin( ) 0
0 ( 0) since [ ] = 1 ; 0
using L'Hopital's rule lim
o
o
j N k k Nj j j l N No
o
o
j jo o
N k kk k le e e e
N k k l Nk k l
Ne k k l X k e
l 0
sin[ ] cos[ ]limsin[ ] cos[ ]
l l Nl N l N
N
![Page 27: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/27.jpg)
BME 310 Biomedical Computing -J.Schesser
342
Two Cases for our ExampleCase 2: when the fundamental frequency of our signal is a multiple of ,
which is the resolution of the sampled spectrum obtained from the DFT.2 (2)ˆthat is say 2 and 40; 2 0.1
40 20If we
s
o o
fN
k N
ˆdo this, the figure below shows that we have a single component at or at 2.ˆIn order words: [ ] ( 2) which corresponds to 2 2 / 40 0.1
o oj
o
k k
X k Ne k
-20
0
20
40
-1 -0.8 -0.5 -0.3 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3
![Page 28: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/28.jpg)
Time Domain & Frequency Domain
• Time Domain
BME 310 Biomedical Computing -J.Schesser
343
LTI SystemCT: Differential EqDT: Difference Eq
CT: ay(t) by(t) cy(t) d x(t) ex(t)
DT: y[n] bk x[n k]k0
M
Input Time signalCT: x(t)DT: x[n]
Output Time SignalCT: y(t)DT: y[n]
![Page 29: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/29.jpg)
Time Domain & Frequency Domain
Time Domain • Impulse Response
BME 310 Biomedical Computing -J.Schesser
344
LTI SystemCT: Differential EqDT: Difference Eq
CT: ay(t) by(t) cy(t) d x(t) ex(t)
DT: y[n] bk x[n k]k0
M
Input Impulse signalCT: δ(t)DT: δ[n]
Output Impulse ResponseCT: h(t)DT: h[n]
![Page 30: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/30.jpg)
Time Domain & Frequency Domain
• Time Domain
BME 310 Biomedical Computing -J.Schesser
345
LTI SystemCT: Differential EqDT: Difference Eq
CT: ay(t) by(t) cy(t) d x(t) ex(t)
DT: y[n] bk x[n k]k0
M
Input Time signalCT: x(t)DT: x[n]
Output Time signalCT: y(t)DT: y[n]
ConvolutionCT: ( ) ( ) ( )
DT: [ ] [ ] [ ]l
y t x h t d
y n x l h n l
![Page 31: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/31.jpg)
Time Domain & Frequency Domain
• Frequency Domain
BME 310 Biomedical Computing -J.Schesser
346
LTI SystemFrequency Response
Input Signal Spectrum
Output Signal Spectrum
CT: H ( j ) h(t)e j t dt
M ( )e j ( )
DT: H (e j ) bkk0
M
e jk h[k]k0
M
e jk
M ( )e j ( )CT: x(t) A0 Ak cos(2 fkt k )
k0
M
DT: x[n] A0 Ak cos( kTsnk )k0
M
A0 Ak cos( knk )k0
M
CT: y(t) A0M (0)e j (0) Ak M (2 fk )cos(2 fkt k (2 fk )k0
M
)
DT: y[n] A0M (0)e j (0) Ak M ( k )cos( kn k ( k ))k0
M
![Page 32: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/32.jpg)
Fourier Transform Signals and SystemsSignals
• The FT of a signal transforms it from the time domain, x(t) or x[n], to the frequency domain to yield its spectral representation, X(jω) or X[k].
• The inverse FT transforms the signal’s spectrum, X(jω) or X[k], in the frequency domain to the time domain, x(t) or x[n].
• The FT can also be applied to systems.
Systems• The FT of a system’s impulse
response, h(t) or h[n], transforms it into the frequency response, H(jω) or H[k].
• The inverse FT transforms a system’s frequency response, H(jω) or H[k] to the impulse response, h(t) or h[n].
• Note that the impulse response is really the output signal of a system when the input signal is the impulse function, δ(t) or δ[n].
BME 310 Biomedical Computing -J.Schesser
347
![Page 33: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/33.jpg)
Fourier Transform Signals and SystemsThe Math is the Same
Signals• For continuous signals, its spectral
representation is:
• For discrete signals, its spectral representation is discrete frequency function of k and is calculated as:
Systems• For systems which process
continuous signals, the frequency response is calculated as:
• For systems which process discrete signals, the frequency response is a continuous frequency function of and is calculated as:
BME 310 Biomedical Computing -J.Schesser
348
H ( j ) h(t)e jt dt
H ( j ) H (e j ) h[k]e jk
k0
M1
bke jk
k0
M1
X( j ) x(t)e j t dt
X[k] x[n]e j 2k
Nn
n0
L1
![Page 34: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/34.jpg)
Fourier Transform Signals and SystemsThe Math is the Same
BME 310 Biomedical Computing -J.Schesser
349
Frequency Response is a continuous function
H (e j ) bke jk
k0
M1
bk {1, 0, 0,1}
H (e j ) bke jk
k0
3
1 0 0 e j 3 1 e j 3
e j3
2 (e j3
2 ej 3
2 ) 2cos(32
)e j3
2
![Page 35: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/35.jpg)
Fourier Transform Signals and SystemsThe Math is the Same
BME 310 Biomedical Computing -J.Schesser
350
Signal Spectral Density is a discrete function
X[k] x[n]e j 2k
Nn
n0
L1
x[n] {1, 0, 0,1}
X[k] x[n]e j 2k
4n
n0
3
1 0 0 e jk
231 e
j3k2
X[k] 1 e j3k
2
X[0] 1 e j 3
20 2
X[1] 1 e j 3
211 j 2e
j4
X[2] 1 e j 3
2211 0
X[3] 1 e j3
231 e
j2 1 j 2e
j4
2 33 34 2 2
0
3 3 3 34 4 4 4
34
3 04
3 3 34 4 4 4
3
OR
[ ] [ ] 1 0 0 1
3( ) 2cos( )4
3[ ] 2cos( )4
3 0[0] 2cos( ) 24
3[1] 2cos( ) 2 2 243[2] 2cos( )2
k k kj n j j
n
k k k kj j j j
kj
j
j j j jj
j
X k x n e e e
ke e e e
kX k e
X e
X e e e e e
X e
2
94 4 4
0
9[3] 2cos( ) 2cos( ) 24 4
j j jX e e e
0 1 2 3 k
X[k]
![Page 36: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/36.jpg)
BME 310 Biomedical Computing -J.Schesser
351
Time Domain to Frequency Domain Transformations
Time DomainTransformation Type
Frequency DomainTime Signal
TypeComputer
processing & storage
Frequency Spectrum
Type
Computer processing &
storage
Periodic & Continuous
x(t)No
Fourier Series DiscreteSpectrum
ak
Yes
Non-Periodic & Continuous
x(t)No
Continuous Time Fourier TransformCTFT
Continuous Spectral Density
X(jω)No
Discrete
x[n]Yes
Discrete Time Fourier TransformDTFT
Continuous Spectral Density No
Discrete
x[n]Yes
Discrete Fourier TransformDFT
DiscreteSpectrum
X[k]
Yes
ak 1T
x(t)e j 2 k
Tt
T /2
T /2
dt
X( j ) x(t)e j t dt
X(e j ) x[n]n
e jn X(e j )
X[k] x[n]n
e j
2N
nk
![Page 37: ComputingBME 310 Biomedical Computing - J.Schesser 321 Fourier Transform • We can handle non-periodic signals in a similar fashion as periodic signals • That is, we can decompose](https://reader030.fdocuments.us/reader030/viewer/2022040803/5e3f49fe143bdc11867d3001/html5/thumbnails/37.jpg)
BME 310 Biomedical Computing -J.Schesser
352
Homework
• Exercises:– 13.2- 13.4, 13.6 – 13.8
• Problems:– 13.1 Use Matlab to plot the spectrum; submit your
code– 13.4, 13.5, 13.6