See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/300114986

A Novel Frequency-Time Based Approach for the Detection of Characteristic

Waves in Electrocardiogram Signal

Chapter · December 2016

DOI: 10.1007/978-81-322-2728-1_6

CITATIONS READS

3 257

2 authors:

Kiran Kumar Patro P. Rajesh Kumar

Andhra University Andhra University
8 PUBLICATIONS 42 CITATIONS 107 PUBLICATIONS 329 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

PhD work View project

ECG Biometrics View project

All content following this page was uploaded by Kiran Kumar Patro on 08 March 2018.

The user has requested enhancement of the downloaded file.

A Novel Frequency-Time Based Approach
for the Detection of Characteristic Waves
in Electrocardiogram Signal

Kiran Kumar Patro and P. Rajesh Kumar

Abstract ECG is the electrical behavior of heart signal which is used to diagnose
the irregularity of heart activity after visually inspecting the ECG signals but it is
difficult to identify by physician’s naked eye hence an effective computer based
system is needed. One cardiac cycle of ECG signal consists of characteristic waves
P-QRS-T. The amplitudes and intervals values of P-QRS-T segment determine the
functioning of heart of every human. In this paper a novel methodology of
Frequency-Time based approach is used to identify P-QRS-T waves. R-peak
detection is the first step in characteristic waves detection, for identifying R-peak,
wavelet transform (sym4) decomposition method (Frequency domain) is used.
After R-peak detection other characteristic waves are detected by tracing to and fro
from R-peak (Time domain). Standard ECG wave form is taken as base wave form
and detects the waves in the estimated interval. MIT-BIH NSR database is taken
and the methodology is implemented on MATLAB software.

Keywords Electrocardiogram (ECG)

Wavelet transform decomposition




Frequency domain Time domain MIT-BIH NSR Matlab

1 Introduction

The Electrocardiogram signal is a recording of the heart’s electrical activity and

provides valuable clinical information about heart’s performance. The electrical
activity generated by Atria and ventricles by its depolarization and repolarization is
depicted in both magnitude and direction in a graphical manner [1]. It provides

K.K. Patro (&)
P. Rajesh Kumar

Department of Electronics and Communication Engineering, Andhra University
College of Engineering, Visakhapatnam, Andhra Pradesh, India
e-mail: kiranugcjrf@gmail.com
P. Rajesh Kumar
e-mail: rajeshauce@gmail.com

© Springer India 2016 57

S.C. Satapathy et al. (eds.), Microelectronics, Electromagnetics
and Telecommunications, Lecture Notes in Electrical Engineering 372,
DOI 10.1007/978-81-322-2728-1_6
58 K.K. Patro and P. Rajesh Kumar

Fig. 1 Characteristic waves

of ECG

information about the morphology, rhythm, and. heart rate. ECG is not unique it
may vary from one person to other person as known fact that there is difference in
position, size, anatomy of the heart, age, relative body weight, chest configuration
and various other factors.
The electrical activity during the cardiac cycle is characterized by five separate
characteristic waves that are P, Q, R, S and T [1]. The amplitude of P wave is very
low (0.1–0.2 mV) and represent depolarization and contraction of the right and left
Atria. QRS complex having largest voltage deflection (1–1.2 mV) and represents
Depolarization of ventricles. T wave amplitude is also low (0.1–0.3 mV) and
represents ventricles Repolarization. The ECG signal is very sensitive in nature
therefore even the presence of small noise the various characteristics of the original
signal changes. Powerline Interference, Baseline wandering noise and EMG noises
corrupts the ECG during data acquisition, which makes detection of characteristic
waves difficult. So proper filtering is needed [2] for accurate detection.
Generally in ECG the major characteristic point is R-peak. Detection of R-peak
is a crucial task and it is the initial job in ECG signal analysis. After finding R-peak
location other components P, Q, S, T are detected by taking R-peak location as
reference and tracing to and fro from R-peak relative position. Figure 1 shows
normal ECG with all characteristic waves.

2 Frequency-Time Analysis for Characteristic Waves

Detection

R-peak detection is the first step in ECG characteristic waves detection, by taking
R-peak as reference, the other characteristic waves was identified Pan Tompkins
algorithm [3] also used to detect ECG characteristic waves. In this frequency
domain analysis (wavelet transform) is used for detection of R-peak [4] and next
Time domain analysis is used for detection of other characteristic waves.
A Novel Frequency-Time Based Approach for the Detection … 59

Fig. 2 Symlet wavelet transform

A. Wavelet Transform Approach

The Wavelet Transform is a time-scale analysis and used on wide range of appli-
cations, in particular signal compression. Now a day’s wavelet transform is used to
solve problems in Electro cardiology, including compression of data, ventricular
late potentials analysis, and finally ECG characteristic waves detection. Wavelet
transform decomposes the given signal into a number of levels related to signal
frequency components and analyses each level with particular resolution [5].
In this the ECG signal decomposed into 4 levels using Symlet wavelet (sym4)
transform for finding R-peak. In wavelet decomposition it down samples the
original ECG signal, as a result the samples are reduced and QRS complex is
retained. Symlet wavelet transform is the modified version of Daubechies wavelet
with increased symmetry (Fig. 2).
Decomposed signals are noise free signals, and by making a threshold of 60 % of
maximum value. The values which are above threshold [6] are invariably R-peaks.
The decomposed signal can be reconstructed into actual signal by first multiplying
the down sampled signal into 4, so that R-peaks are detected in actual signal.
B. Time domain Analysis
Frequency domain approach is used for the detection of R-Peak only and after time
domain approach is used for other characteristic wave’s detection.
R-R interval can be calculated by

Rlocðn þ 1Þ  RlocðnÞ
TRR ðnÞ ¼ ðsecÞ
fs
fs ¼ Sampling frequency
Rloc ¼ location of Rpeak

For identifying P-wave, a window in time domain is created with time gap limits
from 65 % of R-R interval to 95 % R-R interval which is added to same R-peak
location. In that window the maximum value will represents P-wave.
The Q-wave is identified by choosing minimum value in Time based window
starting from 20 ms before corresponding R-peak. Similarly S-peak is detected by
selecting least value in time based window after R-peak location.
60 K.K. Patro and P. Rajesh Kumar

For identifying T-wave, a window in time domain is created with time gap limits
from 15 % of R-R interval to 55 % R-R interval which is added to same R-peak
location. In that window the maximum value will represents T-wave.
The time domain windows are adaptive because they depend on R-R interval
values.

3 Methodology of Frequency-Time Based Approach

The main aim of this methodology is for accurate detection of Characteristic waves
P-QRS-T of Electrocardiogram. In this Frequency-Time domain approach is
adopted [7]. Finding of R-peak is the initial step of identifying characteristic fea-
tures of ECG which is done in frequency domain (wavelet decomposition) method
[7, 8] and other features are detected by taking R-peak location as reference and
thereby creating windows in time domain [9] (Fig. 3).
Steps wise methodology of proposed Frequency-Time based Approach:
Step 1: Load the ECG records from MIT-BIH NSR database [10, 11]
Step 2: Remove the different type of noises in ECG frequency range (Baseline
drift noise, power line Interference, EMG noise) using cascaded based
digital filters (Fig. 4).
Step 3: For finding R-peak decompose the signal using Wavelet (db4/sym4) at
particular scale [7, 8].
Step 4: After decomposition identify R-peak in ECG signal keeping 60 % of the
signal value as threshold.
Step 5: Reconstruction of ECG signal from decomposed signal find R-peak and
R-location (Rloc)
Step 6: For finding Q-point by finding lowest value in the window range
Rloc-X1 to Rloc-Y1
Where X1 = 50 * ts Y1 = 10 * ts (ts = sampling time)
ts= 1/fs (fs= sampling frequency)
Step 7: For identifying S-point create a window on the right side of R-peak in
time domain in the range of
Rloc + X2 to Rloc + Y2
Where X1 = 5*ts Y1 = 50*ts
ts = 1/fs (fs = sampling frequency)
Step 8: Detect T-point by finding highest value in the window range
Rloc + X3 to Rloc + Y3
Where X3 = 25*ts Y3 = 100*ts
ts = 1/fs (fs = sampling frequency)
A Novel Frequency-Time Based Approach for the Detection … 61

Fig. 3 Frequency-time based

methodology

Step 9: Detect P-Point by finding maximum value in window range

Rloc − X4 to Rloc − Y4
Where X4 = 50*ts Y1 = 100*ts
ts = 1/fs (fs = sampling frequency)
All the windows in time domain are created w.r.t Standard wave form of ECG
shown in Fig. 5.
62 K.K. Patro and P. Rajesh Kumar

Fig. 4 Cascaded FIR filter configuration

Fig. 5 Standard ECG

waveform

4 Results and Discussion

The analysis of Frequency-Time based approach was tested MIT-BIH NSR data-
base [10]. The entire detection techniques developed in MATLAB software [12].
The above Fig. 6 represents original ECG signal taken from MIT-BIH NSR
database.
The above Fig. 7 shows Symlet wavelet transform with 4 level decomposition
(Frequency Domain). It is a down sampling process which reduces samples in each
level) (Fig. 8).
The above Fig. 9 shows R-peak detected 4 level down sampled signal by taking
this signal as reference R-peak is detected in actual signal by reconstructing down
sampled signal (Fig. 10).
A Novel Frequency-Time Based Approach for the Detection … 63

original signal
700
600
500
400
300
200
100
0
-100
-200
0 1000 2000 3000 4000 5000 6000 7000 8000

Fig. 6 Original ECG waveform

1st level approx 2nd level approx

1000 1500

1000
500
500
0
0

-500 -500
0 1000 2000 3000 4000 0 500 1000 1500 2000

3rd level approx 4th level approx

1000 600

400
500
200
0
0

-500 -200
0 500 1000 0 200 400 600

Fig. 7 Wavelet decomposition (sym4) of ECG

Finally the R-peaks are detected in actual signal using Wavelet decomposition
and reconstruction methods.
The above Figs. 11 and 12 represents P-QRS-T detected actual ECG signal from
MIT-BIH NSR database 16265.mat signal [10]. In that R-peaks represents by ‘*’,
P-peaks represents by ‘O’, Q and S peaks are represents by ‘+’ and finally T-peak is
represented by’Δ’.
64 K.K. Patro and P. Rajesh Kumar

detail coeff:level 1 cd1 detail coeff:level 2 cd2

200 500

100 0

0 -500

-100 -1000
0 1000 2000 3000 4000 0 500 1000 1500 2000

detail coeff:level 3 cd3 detail coeff:level 4 cd4

1000 1000

500 500

0 0

-500 -500

-1000 -1000
0 500 1000 0 200 400 600

Fig. 8 Detailed wavelet coefficients (sym4) of ECG

R peaks in downsampled signal

1200

1000

800

600

400

200

-200

-400
0 200 400 600 800 1000 1200 1400 1600 1800 2000

Fig. 9 R-peak detected down sampled ECG

The Characteristic waves detected ECG signal is taken from MIT-BIH Normal
Sinus Rhythm database 16,265 mat signal (7500 samples with Sampling frequency
fs ¼ 128 Hz).
The Tables 1 and 2 represents amplitudes and wave locations (samples) of actual
ECG signal for first 10 samples only. The samples can be changed in time mode by
multiplying the samples with ‘ts’ where ts = 1/fs (fs = sampling frequency).
A Novel Frequency-Time Based Approach for the Detection … 65

detect r peak in actual signal

700

600

500

400

300

200

100

-100

-200
0 1000 2000 3000 4000 5000 6000 7000 8000

Fig. 10 R-peak detected in original ECG

1
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
3500 3600 3700 3800 3900 4000 4100 4200 4300 4400 4500

Fig. 11 P-QRS-T detected in original ECG

R peaks in downsampled signal

1200

1000

800

600

400

200

-200

-400
0 200 400 600 800 1000 1200 1400 1600 1800 2000

Fig. 12 Characteristic waves detected in original ECG signal

66 K.K. Patro and P. Rajesh Kumar

Table 1 Characteristic waves amplitude features in volts

Characteristic waves amplitude features
P-wave Q-wave R-wave S-wave T-wave
amplitude (v) amplitude (v) amplitude (v) amplitude (v) amplitude (v)
0.1483 −0.5898 0.9602 −0.5465 0.1123
0.1456 −0.5783 0.9522 −0.5412 0.1214
0.1412 −0.6117 0.9759 −0.5381 0.1222
0.1372 −0.5900 0.9304 −0.5275 0.1260
0.1405 −0.5705 0.9079 −0.5031 0.1159
0.1276 −0.5452 0.8924 −0.5043 0.1123
0.1412 −0.5548 0.9062 −0.5203 0.1135
0.1399 −0.5769 0.9486 −0.5420 0.1164
0.1304 −0.5763 0.9123 −0.5073 0.1272
0.1423 −0.5773 0.8789 −0.5334 0.1245

Table 2 Characteristic waves locations w.r.t. samples

Characteristic waves locations w.r.t samples
P-wave Q-wave R-wave S-wave T-wave
locations locations locations locations locations
136 149 155 161 174
214 227 232 238 252
291 304 310 316 329
369 382 388 393 407
448 461 467 473 487
527 540 545 551 564
606 619 624 630 644
682 695 701 707 720
760 773 779 784 798
838 851 857 863 877

5 Conclusion

In this paper a novel Time-Frequency based methodology has implemented, in that

for R-peak detection Symlet4 wavelet transform decomposition technique is used
and for other peaks detection Time based window technique is developed. The
information about R-peak and QRS complex is very much useful for ECG analysis,
diagnosis, classification, authentication and identification performance. The
obtained amplitude features, interval features of characteristic waves of ECG are
validated for MIT-BIH NSR database records. The results from MATLAB shows
100 % accurate detection rate for all characteristic waves (P-QRS-T) of ECG.
Future scope includes calculation of all wave intervals, amplitude features for ECG
applications like arrhythmia detection and Bio-metric tool.
 A Novel Frequency-Time Based Approach for the Detection … 67

References

1. R. Acharya, J.S. Suri, J.A.E. Spaan, S.M. Krishnan, Advances in Cardiac Signal Processing.
(Springer, Berlin, 2007), ISBN: 3-540-36674-1
2. S. Banerjee, R. Gupta, M. Mitra, Delineation of ECG characteristic features using multi
resolution wavelet analysis method. Meas. Elsevier 45, 474–487 (2012)
3. J. Pan, W.J. Tompkins, A real time QRS detection algorithm. IEEE Trans. Biomed Engg. 32,
230–236 (1985)
4. Lin, H.-Y., Liang, S.-Y.: Discrete-wavelet transform based noise reduction and R-wave
detection for ECG signals. In: IEEE 15th International Conference on e-Health Networking,
Applications and Services, 2013 (Healthcom 2013)
5. P.S. Addison, Wavelet transforms and the ECG: a re-view. Physiol. Meas. 26(5), 155 (2005)
6. B.U. Kohler, C. Hennig, R. Orglmeister, The principles of software QRS detection. IEEE Eng
Biol. Mag 21, 42–57 (2002)
7. C. Li, C. Zheng, C. Tai: Detection of ECG characteristic waves using wavelet transforms.
IEEE Trans. Biomed. Eng. 42, 21–28 (1995)
8. J.S. Sahambi, S. Tandon, R.K.P. Bhatt, Using wavelet transform for ECG characterization.
IEEE Trans Eng. Med. Biol. 16(1), 77–83 (1997)
9. A.K. Manocha, M. Singh, Automatic delineation of ECG characteristic waves using window
search and multi resolution wavelet transform approach. In: ACEEE International Conference
of Emerging Trends in Engineering and Tech
10. http://www.physionet.org/cgi-bin/atm/ATM—MIT-BIH Normal Sinus RhythmDatabase/
11. http://en.wikipedia.org/wiki/File:SinusRhythmLabels.svg
12. Matlab help: MATLAB MATHWORKS

View publication stats