3.3.1 Synchronized averaging. Problem Propose a time-domain technique to remove random noise given the possibility of acquiring multiple realizations.

3.3.1 Synchronized averaging

Problem

• Propose a time-domain technique to remove random noise given the possibility of acquiring multiple realizations of the signal or event of interest

Solution

• Linear filters fail to perform when the signal and noise spectra overlap.

• Synchronized signal averaging can separate a repetitive signal from noise without distorting the signal

• ECG signals may be filtered by detecting the QRS complexes and using their positions to align the waveforms for synchronized averaging

)()()( nnxny kkk

NnnM

nxM

nyM

M

kk

M

kk

M

kk ,...,2,1;)(

1)(

1)(

1

111

)(nxk

)(nxk

)(nk

represents the kth observation of the original uncorrupted signal

represents the noise in the kth copy of the observed signal

Synchronized average:

21

0

21

0

1

0

])([])([

])(][)([)(

yknyxnx

yknyxnxk

N

n

N

n

N

nxy

X is the template

Y is the ECG signal

x yand are the average of the corresponding signals

Figure 3.12

Figure 3.13

Figure 3.14

12

1

12

1

12

1

121212

222

111

12

1

12

1

12

1

)()(

(n),)(

12~1 ,

kk

kk

kk

ji

ji

kkk

nnxny

nnxny

nxnx

j in

nnxny

nnxny

knnxny

Measured ECG = ECG without interference + interference

average edsynchroniz theof variance theis andmean, theis where},,{

}12

,{}12

,{

}1212

1,0{}0,12

12

1{

}12,0{12

1}0,12{

12

1

12,012

10,12

12

1

0. is and small, very is because ,

12,01212

1012,12

12

1

. and are so andt independenmutually are and because ,

12,1212

112,12

12

1

12

1

12

1

12

1

22

22

22

22

2

22

22

12

1

12

1

12

1

ssss

xx

x

x

x

x

x

jiji

xx

kk

kk

kk

ηηxx

nnxny

For synchronized averaging of order M

The resultant signal has a mean close to x

and a variance equal to

(a standard deviation equal to

Advantage

No spectral content of the signal is lost !