248

|

6 Applications and Methods in Biosignal Processing

ECG

R-peaks

low pass

high pass

derivation

squaring

moving

average

QRS

detection

Fig. 6.21: Basic Principle of Pan-Tompkins-algorithm.

ample, in an implantable pacemaker. It was adapted and modified several times after

publication. The basic principle, however, has remained the same, cf. Figure 6.21.

Pan and Tompkins were using a simple Z80- microprocessor at the time and had to

be careful to use simple algorithms that did not overtax this processor, mainly because

of real-time requirements. Therefore, they could not use common standard filters with

higher filter degrees as well as set coefficients that had to be accurate, for example, up

to the fourth digit after the decimal point. Their goal was to be able to implement the

algorithms using simple values that can be expressed as powers of 2, such as 8 = 2³;

because such coefficients can be represented by simple shift-operations of the binary

memory contents. In addition, the analogue biosignals must also be converted into

digital values, i.e. the analogue values are limited in their spectral range beforehand

so that the sampling theorem is fulfilled. They then have to be sampled and digitised.

Fortunately, the biological signals of the heart are not so high-frequent and do not

place excessive demands on the sampling rate. Here, a sampling frequency of 200 Hz

is sufficient, which could be realised well with A/D- converters even back then. How-

ever, the A/D converters at that time only had a low resolution, and so the developers

had to use 8 bit, i.e. 256 values for their filter coefficients.

Fig. 6.22: Simulation model of Pan-Tompkins-Algorithm for Matlab/Simulink or Scilab/Xcos.