ECG Signal Analysis - ECG Acquisition and Automated Remote Processing - page 26

500 Hz. Use of HT in ECG analysis is described in [ 76 , 77 ]. A unique property of
the transform that the transformed output undergoes a polarity change (i.e., crosses
over x axis) whenever the original wave has a slope reversal point. In a typical
ECG waveform, the slope reversal ideally occurs at the fiducial points. However,
the original waveform is required to be preprocessed to remove the noise for
proper detection. PCA has been used for minimization of feature space data
extracted through DWT decomposition using db4 as mother wavelet [ 78 ]. A short-
time Fourier transform (STFT) along with Gabor filter is utilized for R-peak
detection and feature extraction within the QRS zone [ 79 ]. In [ 80 ], a MLP ANN
has been used for classifying the features of ECG detected through discrete Fourier
transform (DFT), PCA, and DWT.

2.4 Method of ECG Signal Analysis

In this section, a few simple methods of ECG signal analysis is described in the
sequence: ECG preprocessing, QRS detection, and ECG feature extraction. The
data analysis algorithm is applied on a single-lead ECG data from Physionet
database (ptb data and mit-db data). At first, all the R peaks are determined from
the data. We propose three different methods for R-peak detection, viz., (a) dif-
ferential ECG with slope-based criteria, (b) square derivative with amplitude
threshold-based search, and (c) amplitude span with slope-based criteria. Next, the
baseline positions are detected in the TP segment of each beat, and baseline
modulated is corrected by an empirical formula. Finally, each wave peaks and
their respective onset and offset points are detected with respect to R-peak loca-
tions. Finally, the time-plane features are computed. The steps of ECG data
analysis are represented by Fig. 2.4 .
The preprocessing of ECG can be done by any suitable method as described in
the Sect. 2.3.1 . In the proposed feature extraction methods, ECG data from
Physionet has been used. For R-peak detection, MIT-BIH arrhythmia data (mitdb)

ECG data
array

Pre-
processing

Calculation of wave
duration, intervals,
and amplitudes

Baseline
modulation
correction

R-peak
detection

Baseline point
detection

Fiducial point
determination

Fig. 2.4 ECG data analysis steps