6.3 Signals of the Cardiovascular System

|

253

Fig. 6.26: Amount frequency response of the moving window integrator (MA) |AMA(ωT)| according to

Figure 6.22.

after bandpass filtering. This search is quite complicated, as the QRS complex should

be found even in a very disturbed ECG signal. Many subsequent publications on this

problem have attempted to simplify this procedure, especially when it is not to be used

in real time. The original procedure is divided into several sections:

1.

Measurement preparations: Since the rising edge of the output signal after win-

dow integration is related to the location of the QRS complex (cf. Figure 6.28), this

allows the associated locations to be approximated. This preliminary investiga-

tion is carried out exclusively on the output signal of the window integrator (MA).

In Matlab, however, there is still the possibility to determine local maxima in the

output signal PEAKI of the window integrator with the command findpeaks(),

which makes the programming very easy. A minimum distance between the local

maxima can also be specified for this command; because it is not possible, due

to the required recovery time of the heart cells, for two QRS complexes to have

a shorter distance than approximately 200 ms from each other. Two thresholds

are provided: THRESHOLDI1 to further investigate if it is a QRS complex, and

THRESHOLDI2 if the first threshold is not exceeded and decided whether to initi-

ate a reverse search for a QRS complex not found. Then the local maximum must

be greater than the second threshold. In [72] another "training phase" of 2 seconds

is provided to initially determine the threshold values.

2.

Adaptive threshold adjustment: After measurement preparation, further max-

ima are searched for in the output signal PEAKI of the window integrator. The

threshold values are adjusted adaptively. If the current local maximum exceeds

the threshold THRESHOLD_I1, this maximum is interpreted as belonging to a

QRS complex (signal peak), if it lies between the thresholds THRESHOLD_I1 and

THRESHOLD_I2, it is classified as a maximum in the noise signal (noise peak),

which initiates a later reverse search. Here, not the local maxima in the signal

PEAKI are chosen as a measure for the size of the QRS complex, but estimated