Hello,

I would like to detect heart beat from ECG signal (Signal I have is 1.5 volts for QRS).
What I would like to do is to measure delta time between each pulsation with enough accuracy (1 milli second would be great). I can use a microcontrolleur with and ADC to do that, but my microcontrolleur as already a lot of think to do ! The best would be to have a pulse when QRS pulse is at the maximum.
I read that an AOP can do that : http://kudelsko.free.fr/articles/aop21.htm#6

But applied to my application, it not so good : too early detection, or too late... what can I do to improve the system ?

Thanks you for your help
Jean-Marie

 

So you are saying the "R" of the QRS peaks at about 1.5 volts?

I would like to detect heart beat from ECG signal (Signal I have is 1.5 volts for QRS).

If you want to do signal conditioning of the QRS pulse and measure the time interval between QRS pulses possibly using a comparator circuit would give you the result you seem to want. That will give you nice square or rectangular pulses to measure the time interval between.

Personally if I wanted to measure ECG signals I would use any of the DAQ (Data Acquisition) inexpensive units available from a number of manufacturers. Most include recording and charting software so you can record the actual ECG signal, then view it and actually measure time intervals between any two points. However, if you just want interval numbers I would use a microcontroller as mentioned and do the signal conditioning using a comparator circuit.

Ron

 

Its in French, is there an English version,

 

You could use a peak detector to establish the "average" level of the R peak. Now attenuate that level by say 10% and use a comparator to compare this level with the actual signal. The comparator should now trigger on each R and let you measure the heart beat frequency.

 

Its in French, is there an English version,

You can use Google Translator. I tried it and while it works, actually works well, it takes awhile.

Ron

 

Hello,

Here is a link with the google translation:
https://translate.google.com/transl...elsko.free.fr/articles/aop21.htm#6&edit-text=

Bertus

 

On many typical EKG graphs, the big divisions are 200 mS. The R peaks are sharp but vary a little in height. Picking the middle of the comparator transitions might narrow the uncertainty of the time estimate sufficiently for a normal rhythm, but whether that reaches the 1-mS criterion would need to be determined by testing.

John

 

Thanks all for your replies,

Its in French, is there an English version,

Sorry Dodgydave... I was only to show the principle. Next time I will take time to found an English site !

You could use a peak detector to establish the "average" level of the R peak. Now attenuate that level by say 10% and use a comparator to compare this level with the actual signal. The comparator should now trigger on each R and let you measure the heart beat frequency.

On many typical EKG graphs, the big divisions are 200 mS. The R peaks are sharp but vary a little in height. Picking the middle of the comparator transitions might narrow the uncertainty of the time estimate sufficiently for a normal rhythm, but whether that reaches the 1-mS criterion would need to be determined by testing.

John

I will try this solution : that will help to lower error coming from signal strength. I will let you know the result when I have some thing working

Jean-Marie

 

Hello,

I hva an issu which I did not see previously : ECG signal present a drift from the moment I put the power one till 20 or 25 secondes where the signal baseline is almost at 0 volt. here two pictures of the drift :

ECG is joint to this message. Where this negative tension could come from ?

Thanks you for your help
Jean-Marie

 

I had issue with picture. Here again :

 

What did you use to measure that?

 

I am using an oscilloscope.

Jean-Marie

 

Hello,

Did that gradual baseline drift come from charging the high-pass filter ?
Or some think like that ?

Thanks you for your help
Jean-Marie

 

I think the C7 R6 390k-10u high pass is the culprit,having time constant of 3.9 seconds. After five time constants (20s) the output should be almost stabilized, which seems to correspond with your findings. Try 1uf and 100k instead, that should give you time constant of 100ms and cutoff point at 1.6hz.
Also C6 might have some effect too.