12.01.2015 Views

Download - Academy Publisher

Download - Academy Publisher

Download - Academy Publisher

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

ISBN 978-952-5726-09-1 (Print)<br />

Proceedings of the Second International Symposium on Networking and Network Security (ISNNS ’10)<br />

Jinggangshan, P. R. China, 2-4, April. 2010, pp. 246-249<br />

ECG Pattern Recognition Based on Wavelet<br />

Transform and BP Neural Network<br />

Shanxiao Yang, and Guangying Yang<br />

School of Physical & Electronics Engineering, Taizhou University, Taizhou, China<br />

ysxtzc@126.com<br />

Abstract—This paper introduces the Electrocardiogram<br />

(ECG) pattern recognition method based on wavelet<br />

transform and standard BP neural network classifier.<br />

Experiment analyzes wavelet transform of ECG to extract<br />

the maximum wavelet coefficients of multi-scale firstly. We<br />

then input them to BP classify for different kinds ECG. The<br />

experimental result shows that the standard BP neural<br />

network classifier’s overall pattern recognition rate is well.<br />

The ECG in this paper was from MIT-BIH. Experimental<br />

result shows that feature vector extracted by the wavelet<br />

transform can characterize ECG patterns, and BP neural<br />

network classifier has a stronger ECG recognition effect.<br />

Index Terms—Electrocardiogram (ECG); Pattern<br />

Recognition; BP neural network; Wavelet Transform;<br />

MIT-BIH<br />

I. INTRODUCTION<br />

Since 1903 the Electrocardiogram(ECG) was<br />

introduced to clinical medicine, the techniques have been<br />

developed rapidly in the record, processing and diagnosis<br />

of the ECG whether it is in the biomedical area or in<br />

engineering and we accumulate considerable experience.<br />

ECG plays an important role in the clinical diagnosis of<br />

the heart disease. It provides an objective indicator for<br />

correct analysis, diagnosis, treatment and care of the heart<br />

disease. Because of its important social value and<br />

economic value, it has a very wide range of applications<br />

in the modern medicine. ECG is still a major research<br />

subject in the biomedical engineering [1].<br />

ECG is the recording of the electrical activity of the<br />

heart, and has become one of the most important tools in<br />

the diagnosis of heart diseases [2]. ECG signal is shaped<br />

by P wave, QRS complex, and T wave. In the normal<br />

ECG beat, the main parameters including shape, duration,<br />

R-R interval and relationship between P wave, QRS<br />

complex, and T wave components are inspected. Any<br />

change in these parameters indicates an illness of the<br />

heart.<br />

The study of ECG recognition has an important<br />

significance in understanding human hear in the role of<br />

human intelligence. Although many efforts have been<br />

taken recently to recognize ECG using different methods,<br />

current recognition systems are not yet advanced enough<br />

to be used in realistic applications.<br />

This work is supported by education department Program of Zhejiang<br />

Province in University (2010).<br />

II.<br />

THE SOURCE OF THE ECG<br />

A. Data Acquisition from the Acquisition Card<br />

Human ECG is a non-linear, non-stationary, weak<br />

signal and low signal to noise ratio. The voltage of the<br />

conventional ECG is just mv level and its frequency is<br />

0.05-100HZ.<br />

Due to equipment, human factors, the acquisition of<br />

ECG accompany with interference. Common ECG<br />

interference has three types 1Electrocardiogram (ECG)<br />

interference. It is caused by human activities and muscle<br />

tension. Its frequency is 5-2000HZ. ECG has wide<br />

spectrum and often mixed with the ECG’s spectrum. So it<br />

is difficult to use the general filtering to separate them. 2<br />

The frequency interference of the power-line. It is fixed<br />

frequency interference. It is form of 50HZ and its<br />

harmonic components. It is caused by the space<br />

electromagnetic interference to human body from powersupply<br />

network and its equipment .In essence, the<br />

alternating current are non-stationary random process in<br />

the frequency and amplitude with a slow fluctuation. 3<br />

Baseline drift. It is caused by the low-frequency<br />

interference, such as the movement of the measuring<br />

electric class, respiratory of the human. Its general<br />

frequency is less than 1HZ [3]. Therefore, the collected<br />

ECG is often accompanied with the above noise.<br />

B. Recognition Methods --BP neural network<br />

BP neural network is fully named as the Back-<br />

Propagation Network, that is, back-propagation network.<br />

It is a forward multi-layer network, which uses the error<br />

back-propagation algorithm to train the network. BP<br />

algorithm [4,5] was proposed by Rumelhart et al in 1986,<br />

and since then, due to simple structure, multi-adjustable<br />

parameters, much training algorithm and good<br />

operational performance, BP neural network got a wide<br />

range of practical application.<br />

The network structure of the three-layer BP neural<br />

network is shown in Figure 1, from which we can see that,<br />

BP neural network contains an input layer, a middle layer<br />

(hidden layer) and an output layer. There is a full<br />

connectivity between the upper and lower layers and no<br />

connections between neurons in each layer. For the input<br />

signal, it needs to spread towards to hidden layer nodes<br />

and transformed by the function, then transmit the input<br />

signal of hidden layer nodes to the output layer nodes.<br />

Usually, the transfer function of BP neural network is<br />

Sigmoid Type differentiable function, which can achieve<br />

arbitrary non-linear mapping between the input and<br />

© 2010 ACADEMY PUBLISHER<br />

AP-PROC-CS-10CN006<br />

246

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!