![](bg5.png)
records from ECG. It is composed of 48 ECG
recording from modified limb lead II (MLII) leads
which are thirty minutes each, and it has over 20
different types of ECG pulses. It has been
constructed a reduced database with 22 records from
these 48 registers. Every beat type is composed by
750 beats. Half of these 750 beats will be used for
training phase and another half for the verification
phase. The detector only detects 8 types of pulse,
which appear with more frequency in this database.
For our experiments, each ANN parameter has
been analyzed to determine if the sign bit is needed,
the minimum number of bits of the integer part for
covering the range and the minimum number of bits
of the decimal. After that the ANN architecture has
been defined and initialized System Generator
simulations can be performed. In these simulations
(fixed point input values) the 98.25% accuracy is
reached with only 1.75% of error (see Table 1).
Table 1: Results of Complete Designed System on Fixed
Point Results.
The chosen FPGA device has been Virtex 7 of
Xilinx and it has been implemented in VHDL. This
implementation in Xilinx allows timming simulation
and its delay is 158.23 ns (much lower than Matlab
delay, about 6.56 ms). The ISE software provides a
power estimator that indicates an estimated power
consumption of 9.3 watts and an estimated
temperature of 42 centigrade degrees in the FPGA.
The occupation in the FPGA is about 45.2% for
logical resources and 39.5% for input-output pins.
5 CONCLUSIONS
In this work, a prototype has been developed on
FPGA, after to observe the conditions of data
dimensions for this device are flexible and highly
parallel architectures, and therefore, it can be built.
These are the reasons that FPGA is suitable for
Digital Signal Processing (DSP). As well as, optimal
results have been reached, the automatic cardiac
pathology detector has a 98.63% of success and the
behaviour simulation of FPGA has been tested in
System Generator with excellent results.
For a close future, this design maybe can be
included in a prototype detector for heart disease
that has elements of communication via infrared,
bluetooth or wi-fi. These radiofrequency signals can
be sent with result from the detection of ECG beat,
so the doctor can have an objective diagnosis in real
time and accurate results on his PDA or personal
computer.
ACKNOWLEDGEMENTS
This work has supported by Cátedra Telefónica –
ULPGC 2010, under the reference, ARUCAS_2010.
REFERENCES
WHO, 2011.World Health Organization. Cardiovascular
Diseases (CVDs). Last view: 07/17/2011.
http://www.who.int/mediacentre/factsheets/fs317.
Lee, D., Bhardwaj, S., Alasaarela, E., Wan Young, Ch.,
2007. An ECG Analysis on Sensor Node for Reducing
Traffic Overload in u-Healthcare with Wireless Sensor
Network.In IEEE SENSOR 2007 Conference.IEEE.
Gu-Young, J., Kee-Ho, Y., 2007.Development of
Ambulatory ECG Monitoring Device with ST Shape
Classification. In ICCAS’07, International Conference
on Control, Automatic and Systems.IEEE.
MIT-BIH DB, 2011. Massachusetts Institute of
Technology –Beth Israel Hospital Database.
Last View: 07/16/2011. http://physionet.org/physiobank/
database/mitdb/
Chen, T., Zheng, Y., Han, L., Guao, P., He, X., 2008. The
Sorting Method of ECG Signlas Based on Neural
Network. In ICBBE 2008, 2
nd
International
Conference on Bioinformatics and Biomedical
Engineering.IEEE.
Pérez, S., Vásquez, J., Ticay, J., Alonso, J., Travieso, C.,
2011. Artificial Neural Network in FPGA for Pejibaye
Palm Classification using Molecular Markers. In
EUROCAST 2011, 13
th
International Conference on
Computer Aided System Theory.IUCTC.
Xilinx, 2011. http://www.xilinx.com/tools/sysgen.htm.
MathWorks, 2011. Last view: 09/01/2011.
http://www.mathworks.es/index.html
Pardo, F., Boluda, J., 2004. VHDL.Language for synthesis
and model circuits. Ra-Ma Editorial, 2
nd
edition.
Ballesteros, D., 2004. FIR Filter Design on FPGA for
Wavelet Background Noise Removal in Bioelectric
Signals. Last view: 08/03/2011. http://redalyc.
uaemex.mx.
BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing
364