Sei sulla pagina 1di 27

Time Frequency Analysis and

Wavelet Transform Tutorial


Time-Frequency Analysis for Biomedical
Engineering

Chia-Jung Chang ()
National Taiwan University

ABSTRACT
Biomedical related research requires lots of mathematical and engineering
techniques to analyze data. Among the subfields, electrophysiological research
plays the core role. In this tutorial, several tools are examined, including
electrocardiogram, and electroencephalogram. These are the most common
tools used to diagnose our physiological activities since neural responses carry
information. Because biomedical signals are usually non stationary, Fourier
transform is not suitable to apply here. Besides, traditional signals are analyzed
in frequency domain, separately from time domain, such that extraordinary
conditions are hard to be observed. To solve such problem, time-frequency
analysis and wavelet transform provide both time and frequency information
simultaneously.
In the following tutorial, I would like to talk about the theoretical background
of both time-frequency analysis and wavelet transform methods, including
what properties they have, their common types, and how to operate them.
Secondly, I would briefly introduce three common physiological tools such as
ECG, and EEG. Where they can be applied, what they are targeting, and what
analysis methods can be used, and how they perform will be described.

CONTENTS
Abstract
I. Theoretical Background
1. Time-Frequency Analysis Methods
1.1 Cohens Class Distribution
1.1.1 Four Derivative Distributions
1.1.2 Electrophysiological Applications
2. Wavelet Transform
2.1 From Fourier to Wavelet Transform
2.1.1 Fourier Transform
2.1.2 Short Time Fourier Transform
2.1.3 Wavelet Transform
2.2 Continuous Wavelet Transform
2.2.1 Properties
2.2.2 Representative Signals
2.3 Discrete Wavelet Transform
2.3.1 Properties
2.3.2 Representative Signals
2.4 Selection of Base Wavelet for Biomedical Signals
2.4.1 Overview
2.4.2 Selection Criteria

II. Biomedical Applications


3. Electrocardiography (ECG)
3.1 Introduction
3.2 Method
3.3 Result
4. Electroencephalography (EEG)
4.1 Introduction
4.2 Method
4.3 Result

Summary
References

1. Time-Frequency Analysis Methods


Great progress has been made in applying linear time-invariant techniques in
signal processing. In such cases the deterministic part of the signal is assumed
to be composed of complex exponentials, the solutions to linear time invariant
differential equations. However, many biomedical signals do not meet these
assumptions. Thus, the emerging techniques of time-frequency analysis can
provide new insights into the nature of biological signals.
There are several different time frequency analysis methods such as short time
Fourier transform, S transform, Wigner transform, and Cohens distribution. I
would focus on Cohens class distribution since it is used in biomedical signals
analysis more commonly than S transform. As for short time Fourier transform,
it is also a popular analysis tool, which will be described in details in the latter
chapters.
1.1 Cohens Class Distribution
Cohens class distribution is defined as follows

1
C x ( t , w )= 2 A x ( , ) ( , ) ei ( t +w ) dd
4

where Ax is called ambiguity function of x(t)

( 2 ) x (t 2 ) e

A x ( , )= x t+

it

dt

Cohens class distribution can be regarded as a series of methods to eliminate


cross-term produced by Wigner distribution, and to keep clarity. However, the
trade-off is that it takes large computation, and thus losses some properties. As
there are different ways to filter out cross-term, there are several forms.
1.1.1

Four Derivative Distributions

There are four different distributions being introduced in this section.


Choi-Williamns distribution (CWD) is defined with
2

( , )=e

()

Born-Jordan distribution (BJ) is defined with


sin
( , )=

( 2 )

Zhao-Atlas-Marks distribution (ZAM) is defined with


( , )=g ( )||

sin( )

While Cohens class distribution can filter cross-term effectively, resolution on


the time-frequency domain is lost at the same time. Another distribution has
been brought out, which is called Reduced Interference distribution (RID). It
can be regarded as smoothed Wigner distribution. After we use a lowpass filter,
the cross term can be easily eliminated, and the resolution is enhanced as well.
Reduced Interference distribution (RID) has been defined as

RID ( t , w ) = h ( ) R x (t , )e iw d

| |
2

Rx (t , )=
||
2

g(v)
2 v

1+ cos
x t +v + x t +v dv

2
2
||

)(

) (

where h(t) is a time smooth window, g(v) is a frequency smooth window.


1.1.2 Electrophysiological Applications
A specific subset of time-frequency distribution is Cohen's class distribution.
For these distributions, a time shift in the signal is reflected as an equivalent
time shift in the time-frequency distribution, and a shift in the frequency of the
signal is reflected as an equivalent frequency shift in the time-frequency
distribution. The Wigner distribution, the RID, and ZAM all have this property.
Tracing back to history, Kawabata employed the instantaneous power
spectrum, a measure of the rate of change of the energy spectrum, to study
dynamic variations in the EEG during photic alpha blocking and performance
of mental tasks. DeWeerd compared time-frequency analysis of event related
potentials obtained by a non uniform filter and the complex energy distribution
function. The applicability of the Wigner distribution for the representation of
event related potentials was examined by Morgan and Gevins.
The RID with a product kernel exhibits the interesting and valuable property of
scale invariance. In addition, the RID has information invariance. Cohen's class
distribution does not vary with time or frequency automatically. On the other
hand, the continuous wavelet transform exhibits scale invariance and time-shift
invariance. In Figure 1a, a spectrogram with a 256-point window is used. In
Figure 1b, a spectrogram with a 32-point window is used. Finally, in Figure 1c,
an RID is used. Spectrogram and RID are compared, and the first is the original
EEG segment. The second is the original EEG segment time-scaled to preserve
energy. The third segment is a frequency-shifted version of the original EEG
segment.

Figure 1
The RID provides a high-resolution representation of both time and frequency.
Furthermore, the RID preserves this structure well even when the original
signal is scaled or shifted in frequency. It is clear that these techniques can be
applied in neurophysiological research such as EEG analysis.

2. Wavelet Transforms
To extract information from signals and reveal the underlying dynamics that
corresponds to the signals, proper signal processing technique is needed.
Typically, the process of signal processing transforms a time domain signal into
another domain since the characteristic information embedded within the time
domain is not readily observable in its original form. Mathematically, this can
be achieved by representing the time domain signal as a series of coefficients,
based on a comparison between the signal x(t) and template functions {n(t)}.

c n= x(t ) n (t) dt

The inner product between the two functions x(t) and n(t) is

x , n = x (t) n (t) dt
The inner product describes an operation of comparing the similarity between
the signal and the template function, i.e. the degree of closeness between the
two functions. This is realized by observing the similarities between the
wavelet transform and other commonly used techniques, in terms of the choice
of the template functions. A non stationary signal is shown in Figure 2 as an
example. The signal consists of four groups of impulsive signal trains. In these
groups, the signals are composed of two major frequencies, 650 and 1500 Hz.

Figure 2
2.1 From Fourier to Wavelet Transform
2.1.1 Fourier Transform
7

Using the notation of inner product, the Fourier transform of a signal can be
expressed as

X ( f )= x , e i 2 ft = x (t)ei 2 ft dt

Assuming that the signal has finite energy

x ( t )= X (f ) e i 2 ft df

The Fourier transform is essentially a convolution between the time series x(t)
and a series of sine and cosine functions that can be viewed as template
functions. The operation measures the similarity between x(t) and the template
functions, and expresses the average frequency information during the entire
period of the signal analyzed as shown in Figure 3.

Figure 3
However, it does not reveal how the signals frequency contents vary with time;
that is, it does not reveal if two frequency components are present continuously
throughout the time of observation or only at certain intervals, as is implicitly
shown in the time-domain representation. Because the temporal structure of the
signal is not revealed, the merit of the Fourier transform is limited; it is not
suited for analyzing non stationary signals.
2.1.2 Short Time Fourier Transform
In Figure 4, the STFT employs a sliding window function g(t). A time-localized
Fourier transform performed on the signal within the window. Subsequently,
the window is removed along the time, and another transform is performed.
The signal segment within the window function is assumed to be stationary. As
a result, the STFT decomposes a time signal into a 2D time-frequency domain,
and variations of the frequency within the window function are revealed.

Figure 4
STFT can be expressed as
i 2 ft

STFT ( , f )= x , gt ,f = x ( t ) gt , f ( t ) dt= x ( t ) g(t ) e

dt

According to the uncertainty principle, the time and frequency resolutions of


the STFT technique cannot be chosen arbitrarily at the same time.
1
4
2
2|g ( )| d

f
2

|g ( )| d

2
f |G(f )| df

=
2
|G(f )| df

As shown in Figure 5, the products of the time and frequency resolutions of the
window function (i.e., the area of f) are the same regardless of the window
size ( or 0.5).

Figure 5
2.1.3 Wavelet Transform
Wavelet transform is a tool that converts a signal into a different form. This
conversion reveals the characteristics hidden in the original signal. The wavelet
is a small wave that has an oscillating wavelike characteristic and has its energy
concentrated in time.
The first reference to the wavelet goes back to the early twentieth century.
9

Harrs research on orthogonal systems of functions led to the development of a


set of rectangular basis functions. The Haar wavelet (Figure 6) was named on
the basis of this set of functions, and it is also the simplest wavelet family
developed till this date.

Figure 6
In contrast to STFT, the wavelet transform enables variable window sizes in
analyzing different frequency components within a signal. By comparing the
signal with a set of functions obtained from the scaling and shift of a base
wavelet, it is realized as shown in Figure 7.

Figure 7
Wavelet transform can be expressed as

wt ( s . )= x , s , =

1
t
x (t )
dt

s
s

( )

As in Figure 8, variations of the time and frequency resolutions of the Morlet


wavelet at two locations on the timefrequency plane, (1, s1) and (2, s2)
are illustrated.

10

Figure 8
Through variations of the scale and time shifts of the base wavelet function, the
wavelet transform can extract the components within over its entire spectrum,
by using small scales for decomposing high frequency parts and large scales for
low frequency components analysis.
2.2 Continuous Wavelet Transform
2.2.1 Properties
The definition of continuous wavelet transform

X (a , b)=

1
ta
x (t )
dt

b
b

( )

where a shifts time, b modulates the width (not frequency), and(t) is mother
wavelt.
It has superposition property.
If the continuous wavelet transform of x(t) is X (s,) and of y(t) is Y(s,), then
the continuous wavelet transform of z(t) = k1x(t) + k2y(t) can be expressed as
Z ( s , )=k 1 X ( s , ) +k 2 Y ( s , )

Moreover, it is covariant under translation and dilation.


Suppose that the continuous wavelet transform of x(t) is X (s,), then the
transform of x(t-t0) is
X (s , t 0)

This means that the wavelet coefficients of x(t-t0) can be obtained by


translating the wavelet coefficients of x(t) along the time with t0.
On the other hand, suppose that the continuous wavelet transform of x(t) is X
(s,), then the continuous wavelet transform of x(t/a) can be expressed as
11

a X

( as , a )

This indicates that, when a signal is dilated by a, its corresponding wavelet


coefficients are also dilated by a along the scale and time axes.
2.2.2 Representative Signals
There are several commonly used wavelets for performing the CWT.
One is Maxican hat, which is a normalized second derivative of a Gaussian
function, and frequently employed to model seismic data
t

1
2 2
(t)=
1 2 e
t
2 3

( )

2
2

Another is Morlet, which has been used to identify transient components


embedded in a signal, bearing defect-induced vibration.
1
(t)=
ei 2 f t e
fb
c

t
fb

Another is Frequency B-Spline Wavelet, which has been seen in biomedical


signal analysis.
fbt
p
sin

( t ) = f b
c

( )

where fb is the bandwidth parameter, fc is the wavelet center frequency, and p is


an integer parameter.
There are also many other mother wavelets such as Shannon Wavelet (a special
case of frequency b-spline wavelet), Gaussian Wavelet, Harmonic wavelet, etc.
2.3 Discrete Wavelet Transform
2.3.1 Properties
It can be implemented if it is 1D as shown in Figure 9, and 2D case is shown in
Figure 10. Similarly, it can be fitted into higher dimensionality.

Figure 9
12

Figure 10
Simply and clearly, we can tell from the implmentation structures that it has
higher computational speed than CWT, and the signals would be continuously
seperated into low frequencies and high frequencies as shown in Figure 11.

Figure 11
2.3.2 Representative Signals
There are several commonly used wavelets for performing the DWT.
One is Haar, which is orthogonal and symmetric. The property of symmetry
ensures that the Haar wavelet has linear phase characteristics, meaning that
when a wavelet ltering is performed on a signal with this base wavelet, there
will be no phase distortion in the ltered signal. Furthermore, it is the simplest
base wavelet with the highest time resolution.

1 0 t <0.5
( t ) = 1 0.5 t <1
0 otherwise

However, the rectangular shape of the Haar wavelet makes its corresponding
spectrum with slow decay, leading to a low frequency resolution.
13

Another is Daubechies, is orthogonal and asymmetric, which introduces a


large phase distortion. This means that it cannot be used in applications where
a phase information needs to be kept. It is also a compact support base wavelet
with a given support width of 2N - 1, in which N is the order of the base
wavelet. 2 base and 4 base Daubechies transforms are illustrated in Figure 12.

Figure 12
The Daubechies wavelets have been widely investigated for fault diagnosis of
bearings and automatic gears.
There are also Coiflet and Symlet, extended from Dauchechies families, but
are more symmetric and have vanishing moments in scaling functions.
2.4 Selection of Base Wavelet for Biomedical Signals
2.4.1 Overview
One of the advantages of wavelet transform for signal analysis is the abundance
of the base wavelets. From such abundance arises a natural question of how to
choose a base wavelet that is best suited for analyzing a specic signal. Since
the choice in the rst place may affect the result of wavelet transform at the
end, the question is valid. For example, as shown in Figure 13,
Therefore, in the following section, we rst present a general strategy for base
wavelet selection. Then, we introduce several quantitative measures that can be
used as guidelines for wavelet selection. While Morlet wavelet is effective in
extracting the impulsive component, the Daubechies and Mexican hat wavelets
did not fully reveal the characteristics of impulsive component.

14

Figure 13
2.4.2 Selection Criteria
There are two ways to measure wavelet performance, one is qualitative and the
other is quantitative.
Base wavelets are characterized by orthogonality, symmetry, and compact
support. Understanding these properties will help choose a candidate base
wavelet from the wavelet families for analyzing a specic signal. For example,
the orthogonality property indicates that the inner product of the base wavelet
is unity with itself, and zero with other scaled and shifted wavelets. As a result,
an orthogonal wavelet is efcient for signal decomposition into nonoverlapping
subfrequency bands. The symmetric property ensures that a base wavelet can
serve as a linear phase lter. A compact support wavelet is one whose basis
function is nonzero only within a nite interval. This allows the wavelet
transform to efciently represent signals that have localized features.
In the area of biomedical engineering, the regularity and symmetry of base
wavelets were considered as essential features for auditory-evoked potentials
(AEP) signal analysis. The morphology and latency of peaks were preserved
with a symmetric base wavelet. By using the properties of compact support,
vanishing moment, and orthogonality, the Coiet wavelet was selected to
separate burst andtonic components in the electromyogram (EMG). In addition
to orthogonality, the property of complex or real basis was used to guide the
choice of the basewavelet for electrocardiogram (ECG) signal analysis. The
Morlet wavelet, Gaussian wavelet, and quadratic B-Spline wavelet were
preselected as the candidates for ECG detection and segmentation.
Besides above properties, shape matching is alternative to wavelet selection.
For example, to measure the timing of multiunit bursts in surface EMG from
single trials, wavelets of different shapes, such as square, triangular, Gaussian
15

and Mexican Hat, were investigated. The Daubechies wavelet was chosen for
its similarity to the shape of motor unit potentials hidden in the EMG signal.
Also, base wavelets of different shapes were compared with ECG signals to
determine their appropriateness for extracting a reference base from corrupted
ECG.
As far as shape matching is concerned, it is generally difcult to accurately
match the shape of a signal to that of a base wavelet through a visual
comparison. These deciencies motivate the study of quantitative measures for
base wavelet selection.
In the area of biomedical engineering, study on horse gait classication has
discussed an uncertainty model for wavelet selection. The model combines
the fuzzy uncertainty with the probabilistic uncertainty to provide a better
measure for choosing base wavelet to improve correct classication of different
horse gait signals.
In study on cardiovascular ailments in patients, experiments have revealed the
suitability of the Daubechies wavelet at order 8 for the ECG signal denoising,
as it has the maximum cross correlation coefcient between the ECG signal
and the chosen base wavelets, (Daubechies, Symlet, and Coiet wavelets).
From above discussion and references, we know that
Coiflect 4 effectively separated burst and tonic components for EMG.
(Wang et al. 2004)
Morlet, Gaussian, Paul 4, and quadratic B-Spline wavelets were selected
for ECG detection and segmentation. (Bhatia et al. 2006)
Symlet 7, Coifect 3, Coifect 4, and Coifect 5 wavelets have better
detection for ECG. (Abi-Abdallah et al. 2006)
Furthermore, there are two more quantitative creteria to determine optimal
wavelet base, which are maximum energy and minnimum Shannon entropy.
The energy content of a signal x(t) can be calculated by
2

E=| x( t)| dt

It can also be calculated from its wavelet coefficients wt(s, )


2

E=|wt ( s , )| dsd

The above equation can be revised as


2

E( s)=|wt ( s , )| d

If a major frequency component corresponding to a specific scaling s exists in


the signal, then the wavelet coefficients at that scale will have relatively high
magnitudes when this major frequency occurs. Thus, the energy related to that
component can be extracted from the signal when applying wavelet transform.
Therefore, we want a base wavelet that can extract the largest amount of energy
16

from the signal.


The entropy here refers to the entropy of the wavelet coefficients, instead of
the one of the signal itself. The energy distribution of the wavelet coefficients
can be described by Shannon entropy.
N

Eentropy ( s ) = p i log 2 pi
i =1

where pi is the energy probability distribution of wavelet coefficients


2

pi=

|wt (s ,i)|

Eentropy ( s )

From the probability principle, we find the bound of the entropy of the wavelet
coefficients.
0 E entropy ( s ) log 2 N

If all the other wavelet coefficients are equal to zero except for one coefficient,
the entropy equals to zero. If the probability distribution is uniform, which
means all the wavelet coefficients are the same (i.e. 1/N), the entropy equals to
log2N. Therefore, the lower the entropy is, the higher energy concentration is.

3. Electrocardiography (ECG)
3.1 Introduction
Electrocardiography (ECG) is a transthoracic interpretation of the electrical
activity of the heart over a period of time, as detected by electrodes attached to
the outer surface of the skin and recorded by a device external to the body. In
short, electrocardiogram is a test that records the electrical activity of the heart.
ECG has been used to measure the rate and regularity of heartbeats as well as
the size and position of the chambers, the presence of any damage to the heart,
and the effects of drugs or devices used to regulate the heart, as shown in
Figure 14 and 15.

Figure 14

17

Figure 15
A typical ECG tracing of the cardiac cycle consists of a P wave, a QRS wave, a
T wave, and a U wave. The baseline voltage of the electrocardiogram is known
as theisoelectric line. Typically the isoelectric line is measured as the portion of
the tracing following the T wave and preceding the next P wave, as shown in
Figure 16 and Table 1.

Figure 16
Feature
RP interval
P wave

PR interval

QRS wave
T wave

Description
The interval between an R wave and the next R wave . Normal
resting heart rate is between 60 and 100 bpm
During normal atrial depolarization, the main electrical vector
is directed from the SA node towards the AV node. This turns
into the P wave on the ECG.
The PR interval is measured from the beginning of the P wave
to the beginning of the QRS wave. The PR interval reflects the
time the electrical impulse takes to travel from the sinus node
through the AV node and entering the ventricles. The PR
interval is therefore a good estimate of AV node function.
The QRS complex reflects the rapid depolarization of the right
and left ventricles. They have a large muscle mass compared to
the atria and so the QRS wave usually has a larger amplitude
than the P wave.
The T wave represents the repolarization of the ventricles. The
interval from the beginning of the QRS wave to the apex of the
T wave is referred to as the absolute refractory period. The last
18

ST interval

half of the T wave is referred to refractory period .


The ST interval is from the J point to the end of the T wave.
Table 1

3.2 Method
We need to select the QRS wave part from ECG to identify if the heart function
is normal or not. The first step is data acquisition with wavelet transform to get
lowpass and hypass signals, as shown in Figure 17.

Figure 17
3.3 Result
The different linear and quadratic approaches of time frequency
representations, such as the spectrogram, the wavelet transform, and the
Wigner distribution, transform a 1D signal x(t) into a 2D function of time and
frequency.
In the left panel of Figure 18, Wigner distribution of (a) a healthy control and
(b) a patient with ventricular tachycardia is compared; in the right panel of
Figure 18, schaologram is used.

19

Figure 18
In Figure 19, ECG has been applied with Morlet Wavlet for the same control
and experimental data.

Figure 19
In short, these time-frequency analysis methods are mainly processed ECG
data as illustrated in Figure 20.

20

Figure 20

4. Electroencephalography (EEG)
4.1 Introduction
Electroencephalography (EEG) is the recording of electrical activity along
the scalp. It measures voltage fluctuations resulting from ionic current flows
within the neurons of the brain. In clinical contexts, EEG refers to the
recording of the brain's spontaneous electrical activity over a short period of
time as recorded from multiple electrodes placed on the scalp. Diagnostic
applications generally focus on the spectral content of EEG, which means the
type of neural oscillations that can be observed in EEG signals. In neurology,
the main diagnostic application of EEG is in the case of epilepsy, as epileptic
activity can create clear abnormalities on a standard EEG study. In figure 21,
several lines of EEG aree illustrated.

Figure 21
To get a closer look, we choose one EEG signal out within one second, as
shown in Figure 22.

Figure 22
It is typically described in terms of rhythmic activity and transients. Then, the
21

activity is divided into bands by frequency, which will have the following
types, from alpha wave, beta wave, delta wave, to theta wave, as shown in
Figure 23.

Figure 23
By analysizing the composition of EEG with wavelet transform, we categorize
and find out the combination percentage of each form.
4.2 Method
Fourier transform, wavelet transforms are used. There is no much complex
algorithms here.
4.3 Result
Suppose there are two EEG data, one is from the patient and the other is from
healthy man, as shown in Figure 24.

22

Figure 24

Among spectral analysis techniques, Fourier transform is considered to be the


best transformation between time and frequency domains because of it being
time-shift invariant, as shown in Figure 25.

Figure 26
If we decompose the EEG signals with wavelet transform, the effect would be
clearer if we want to categorize different subgroups and find the characteristics
of EEG signals between the patient and a healthy man. Also, by using wavelet
transform, one can view the shapes of the subspectral components of the EEG
signal in time domain to be different from those in Fourier transform, as shown
in Figure 27.

23

Figure 27

24

SUMMARY
This tutorial starts from the introduction of theoretical backgrounds of how to
deduce time-frequency analysis methods such as Cohens class distribution,
short time Fourier transform, and wavelet transform. By describing how the
mathematical formulas come out, what properties they have, and how to select
optimal wavelets for our biomedical signal processing, we can gain help from
the brief categories and descriptions.
I didnt talk much on the second part, i.e. ECG and EEG applications, because
the main concept of how to operate these analysis tools have been mentioned in
the first part already, therefore, only brief method descriptions and the result
figures are shown.
There are still many other analysis tools with more complex algorithms that
have not been introduced here. Readers who are interested can find more
information in the references.

REFERENCES
[1] R.X. Gao, R. Yan, Wavelets: Theory and Applications for Manufacturing,
Springer, 2010
[2] M. Akay et al. Time Frequency and Wavelets in Biomedical Signal
Processing, Wiley-IEEE Press, 1997
[3] L. Cohen, "Time-frequency distributions-A review," Proc. IEEE, vol. 77,
pp.941-981, 1989.
[4] L. Cohen, Time-Frequency Signal Analysis. Englewood Cliffs, NJ:
Prentice Hall, 1995.
[5] T.C.M. Claasen and W.F.G. Mecklenbrauker, "The Wigner distribution-A
tool for time-frequency signal analysis-Part I: Continuous-time signals,"
Philips J. Res., vol. 35, pp. 217-250, 1980.
[6] T.C.M. Claasen and W.F.G. Mecklenbrauker, "The Wigner distribution-A
tool for time-frequency signal analysis-Part II: Discrete time signals,"
Philips J. Res., vol. 35, pp. 276-300, 1980.
[7] A. Nayak, R.J. Roy, and A. Sharma,"Time-frequency spectral
representation of the EEG as an aid in the detection of depth of
anesthesia," Ann. Biomed. Eng., vol. 22, pp. 501-513,1994.
[8] T. C.M. Claasen and W.F.G. Mecklenbrauker, "The Wigner distribution-A
tool for time-frequency signal analysis-Part III: Relations with other
time-frequency signal transformations," Philips J. Res., vol. 35, pp. 372389, 1980.
[9] W. J. Williams and T. H. Sang, "Adaptive RID kernels which minimize
25

time-frequency uncertainty," IEEE Int. SY111p. Time-Frequency and


Time-Scale Analysis, pp. 96-99, 1994.
[10] W.J. Williams, "Time-frequency analysis of biological signals," IEEE
Int.Conf. Acoust. Speech, Signal Proc., vol. I, pp. 83-~6, 1993.
[11] C.S. Zheng, W.J. Williams, and J.C. Sackellares, "RID time-frequency
analysis of median filters and lowpass filters in reducing EMG artifacts in
EEGrecording," Proc. IEEE Int. Con. Eng. Med. Bioi., vol. 15, pt. I, pp.
350-351,1993.
[12] Matsuyama, M. Jonkman, F. de Boer, Improved ECG Signal Analysis
Using Wavelet and Feature Extraction, Methods Inf. Med.,vol. 46,
pp.227-230, 2007
[13] Awais M. Kamboh, Matthew Raetz, Karim G. Oweism, and Andrew
Mason, Area-Power Efficient VLSI Implementation of Multichannel
DWT for Data Compression in Implantable Neuroprosthetics, IEEE
Transaction on Biomedical Circuits and Systems, VOL. 1, NO. 2, pp.128135, 2007
[14] C.L. Robert and C.L. William, Physiology of the heart and circulation.
4th. edition. Year Book Medical Publishers, 1989
[15] J. Sylvain , M. Claire , M.B. Pierre, M. Nicola, M. Laurent, V.
Christophe, C. Philippe, M. Alberto, S. Bernard, Instant Power
Spectrum Analysis of Heart Rate Variability During Orthostatic Tilt
Using a Time Frequency Domain Method, Vol. 96, pp.3521-3526, 1997
[16] J.C. Wood, A. J. Buda, and D.T. Barry,"Time-frequency transforms: A
new approach to first heart sound frequency analysis," IEEE Trans. Eng.
Med. Bioi., vol. 38, pp. 728-739, 1992.
[17] E.J. Berbari and R. Lazzara, "The significance of electrocardiographic
late potentials: Predictors of ventricular tachycardia," Annu. Rev. Med.,
vol. 43, pp. 157-169, 1992.
[18] M. Malik, O. Odemuyiwa, J. Poloniecki, P. Kulakowski, T. Farrell, A.
Stauton, and A.J. Camm, "Late potentials after acute myocardial
infarction. Performance of different criteria for the prediction of
arrythmic complications," Eur. Heart. J., vol. 13, pp. 599-607, 1992.
[19] J. Schramm, C. Frumento, H. Dickhaus, C. Schmitt, and J. Brachmann,
"Ventricular late potentials in patients with and without inducible
sustained ventricular tachycardia: A combined analysis in the time and
frequency domain," Eur. Heart. J., vol. 12, pp. 563-572, 1991.
[20] E. J. Berbari, B. J. Scherlag, R. R. Hope, and R. Lazzara, "Recording
from the body surface of arrhythmogenic ventricular activity during the
ST-segment," Am. J. Cardiol., vol. 41, pp. 697-702, 1978.
26

[21] M. B. Simson, "Use of signals in the terminal QRS complex to identify


patients with ventricular tachycardia after myocardial infarction,"
Circulation, vol. 64, pp. 235-242, 1981.
[22] G. Breithardt, M. E. Cain, N. EI-Sherif, N. C. Flowers, V. Hombach, M.
Janse, M.B. Simson, and G. Steinbeck, "Standards for analysis of
ventricular late potentials using high resolution or signal-averaged
electrocardiography," Eur. Heart J., vol. 12, pp. 473-480, 1991.
[23] M. E. Cain, H. D. Ambos, F. X. Witkowski, and B. E. Sobel, "Fast
Fourier transform analysis of signal-averaged electrocardiograms for
identification of patients prone to sustained ventricular tachycardia,"
Circulation, vol. 69, pp.711-720, 1984.
[24] L. Senhadji, J. L. Dillenseger, F. Wendling, C. Rocha, and A. Kinie,
"Wavelet analysis of EEG for three-dimennional mapping of epileptic
events . " Ann. Biomed. Eng., vol. 23, pp. 543-551, 1995.
[25] K. Whittingstall, NK. Logothetis, "Frequency-band coupling in surface
EEG reflects spiking activity in monkey visual cortex" Neuron 64 (2):
2819, 2009
[26] P.L. Nunez, and K.L. Pilgreen, "The spline Laplacian in clinical
neurophysiology: a method to improve EEG spatial resolution". J Clin
Neurophysiol 8 (4): 397413, 1991
M Akin, Comparison of Wavelet Transform and FFT Methods in the
Analysis of EEG Signals Journal of Medical Systems, Vol. 26, No. 3,
2002

[27]

27

Potrebbero piacerti anche