advice on sound analysis of a heartbeat
i need a few tips on the best approach for analyzing the sound of a heartbeat. the sound is grabbed by a microphone hooked to a stethoscope. the results of the analysis will be used to switch light bulbs on and off in an installation.
here is the tricky part : anybody in the audience can hook to the stethoscope in order to trigger the lights, which means a lot of interferences will be caused by the manipulation of the stethoscope. so here is my question : how would you go for separating in a subtle manner the "noise" from the heartbeats ? the thing is that human heartbeats range between 50 and 200Hz and may vary from person to person ( in speed and amplitude ). Manipulating a microphone also causes some noises in the low frequency range.
any help will be much appreciated, thanks in advance
Hey, did you manage to do this project? I would love to hear how you did it as I am trying to do the same for a dance performance.
Much appreciation for your feedback.
What about installing a tilt sensor on the stethoscope? Ignore audio input when the sensor reports movement. Listen to audio input when the sensor reports that the stethoscope is relatively still. (This would mean a lot of listening to "nothing" when no one was in the installation. But isn’t that always the case?) This wouldn’t remove all possibility of handling noise, since someone’s finger could slip or tap the stethoscope while they held it mostly still.
I also wonder if the handling noise would be much louder than the heartbeat signal. This seems to be the case for recordings I’ve made using contact mics. If so, your patch could ignore signal peaks above a certain amplitude.
What you need, I think, is a delay and gate. Ignore signals that spend more than X% of the time above a threshold. Perhaps also ignore signals which pass a second threshold.
thaks for your interest , comments and advices .
@jupiter2260 : experiments are still going on i will put the project on line in a few weeks when i ll have a finished satisfying version of the installation. so far i am using an "arranged" stetoscope into which we added a little mic. the signal is amplified and then analyzed. the analysis needs more work. since it is a participatory installation the difficulty is to make it work in 99% of the cases. i have been surprised to notice that peoples’ hearts really dont beat the same !!! :)
@raja : the thing is that it is an interactive installation so the heartbeat is analyzed in real time, no post processing possible alas, i’ll check Alan’s EKG solution, thanks for the link.
@HollaHopson : thanks for your tilt sensor idea, i will check into that , it can certainly help to get rid of a lot of noise . i like the idea.
@Chris rolfe: filtering and thresholding the signal is indeed a real art in this case . There are of course numerous devices , not sound based but pressure based that analyze heart beats. My collaborator and i find for metaphorical purposes that it is rally about the heart’s "beat", its sound to produce the flickering of the light bulbs. so we kind of stick to the mic-stetoscope combo, but have to do a fine sound analysis job,and as none of us are sound engineers it is quite an "interesting" task :)
thanks again for your inputs , they are very welcome.
best, more soon on the piece .
if you’re going to examine the ecg option, it may be worthwhile to examine a PPG (photoplethysmograph).
uses IR light to register blood flow through a peripheral, ecg can pick up interference from muscle signals fairly severely if not filtered with high tech adaptive filters, PPG has no such electrical muscle interference, meaning the subjects may attach the device and move about without destroying the signal. if in a live environment with "lay" people so to speak, they may not be interested in sitting/lying perfectly still while you process the signal. PPG’s are fairly easy to build, there’s quite a few designs on the internet already,,
Forums > MaxMSP