Biosignal Processing 4
Prof. Jang-Yeon Park
Biosignal Processing 3
Prof. Jang-Yeon Park
Biosignal Processing 1
Prof. Jang-Yeon Park
건국대학교 의학공학부 교수 박장연
Biosignal Processing 5
Biosignal Processing 5
1. The concept of a filter
2. Signal Averaging
- A low-pass filter
- A high-pass filter
- A band-pass filter
1. Filters
Filters are a special class of linear systems:
A filter allows us to selectively remove an undesired signal component while preserving or enhancing some components of interest.
Examples:
a. Sunblock: A type of filter that removes an unwanted ultraviolet light from the sun.
b. Treble and bass controls in an audio system: A type of filter that enables boosting or suppressing the amount of a high
frequency (“treble”) and a low frequency (“bass”) sounds.
The output of a filter is the convolution of the input and the filter’s impulse response.
1. Filters
Filters plays an important role in the analysis of biological systems by removing unwanted noises which distort the signal waveforms and, thus, prevent a reliable diagnosis.
Most filters can be subdivided into three broad classes:
A low-pass filter, A high-pass filter, A band-pass filter.
a. A low-pass filter: It removes the high frequencies from a signal, keeping the low frequencies.
b. A high-pass filter: It passes the high frequencies but removes the low frequencies (“exactly opposite to the low-pass filter).
c. A band-pass filter: It removes both high and low frequencies but selectively keep a small “band” of frequencies.
1. Filters
a. A low-pass filter: It removes the high frequencies from a signal, keeping the low frequencies.
b. A high-pass filter: It passes the high frequencies but removes the low frequencies (“exactly opposite to the low-pass filter).
c. A band-pass filter: It removes both high and low frequencies but selectively keep a small “band” of frequencies.
Low-pass, High-pass, and Band-pass Filters
(Introduction to Biomedical Engineering, by John Enderle et al.:p.591)
An Ideal Low-pass (Analog) Filter
The transfer function of the ideal low-pass filter (b):
c c
LP W
H W
|
| ,
0
|
| ,
) 1
(
frequency cutoff
filter A
f W
where
t W W
t h
c c
c c
LP
: 2
, sinc
) (
The impulse response of the ideal low-pass filter (a):
(Introduction to Biomedical Engineering, by John Enderle et al.:p.591)
An Ideal High-pass (Analog) Filter
The transfer function of the ideal high-pass filter:
c c LP
HP W
H W
H 1, | |
|
| ,
) 0 ( 1
)
(
,sinc )
( )
( )
( )
( W W t
t t
h t
t
hHP LP c c
The impulse response of the ideal high-pass filter:
An Ideal Band-pass (Analog) Filter
The transfer function of the ideal band-pass filter:
filter pass
- low a
of frequency cutoff
the
filter pass
- high a
of frequency cutoff
the
otherwise , ,
0
|
| ,
) 1 ( )
( )
(
2 1
2 1
W W where
W H W
H
HBP HP LP
) ( )
( )
( t h t h t
h
BP
HP
LPThe impulse response of the ideal high-pass filter:
Example 10.21:
An EMG signal contains energy within the frequencies 25 and 100 Hz. Design a filter to remove unwanted nosies:
Hint) We need a “band-pass” filter with pass-band frequencies 25 and 100 Hz.
Wt t
t
W t
t
hHP( ) ( ) 2 1sinc 1 ( ) 50sinc 50
W t
t
W t
hLP( ) 2 2sinc 2 200sinc 200
t
t
t
t h
t h t
hBP( ) LP( ) HP( ) 200sinc 200 ( ) 50sinc 50
The impulse response of the band-pass filter required is:
(Introduction to Biomedical Engineering, by John Enderle et al.:p.593)
2. Signal Averaging
Biological measurements are often confounded by noises.
If the spectrum of noise and signal components do not overlap in the frequency domain, a filter can be easily designed and used to remove the unwanted noises.
However, a filter does not work well in some cases where biological signals and noise spectrums overlap.
In this case, “signal averaging” is a good alternative way to eliminate the noises.
2. Signal Averaging
Many biological signals can be modeled as the sum of an ideal noiseless signal component, x(t), and a separate
independent noise term, n(t):
) ( )
( )
( t x t n t
x
i
x(t): the measured ith trial or ith measurement of the signal.
n(t): a random noise term which causes a trial-to-trial variability.
2. Signal Averaging
) ( )
( )
1 ( )
( )
1 ( )
(
1 1
t t
x t
N n t
x t
N x t
x
N
i N
i
i
If the noise terms, n(t), is purely random, (t) approaches 0 as N . Thus, x(t) x(t) for a very large N where
becomes very small. “A very powerful result!”
Averaged signals:
If we average a sufficiently large number of signal trials, the averaged signal closely approximates the true
noiseless signal waveform.
2. Signal Averaging
(Introduction to Biomedical Engineering, by John Enderle et al.:p.553,p.600)
2. Signal Averaging
(Introduction to Biomedical Engineering, by John Enderle et al.:p.601)