Can this algorithm recognize a person playing a song, e.g. Fur Elise on the piano against a recording , say by the Boston Symphony or just another individual

[D Si]

In other words, I'd like to do some sound recognition but it's not a direct match of a radio station playing a song against that song in a database. It's more along the lines of several individuals making the sound but each may be a but different but we know it's the same sound.

Please advise.

Thx

[Arijit Mallik]

Hi,

I am not sure whether this algorithm works in this case and even if it works, whether the accuracy is high enough. However, you can do a small tweak: train a multi-class machine-learning classifier which puts all the records of a particular song (say, voices and all possible instruments) into one class. When your classifier has high accuracy on your training data-set and cross-validation data set; you can be sure. Be careful about the feature set.

Hope the above makes sense.

Thanks,

Arijit

--
You received this message because you are subscribed to the Google Groups "echoprint" group.
To unsubscribe from this group and stop receiving emails from it, send an email to echoprint+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[D Singh]

Thanks. I am actually trying to classify individual patient auscultation sounds with data gathered from a digital stethoscope.

It's a proof of concept at this stage and there aren't that many patterns that actually need identifying, the catch is it will sound a bit different for each patient.

I know that has been done with heart auscultations before but in this application there are no standards to classify the sounds like there is with heart sounds.

Any recommendations welcome.

--
You received this message because you are subscribed to a topic in the Google Groups "echoprint" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/echoprint/n7LYmRamJ_o/unsubscribe.
To unsubscribe from this group and all its topics, send an email to echoprint+unsubscribe@googlegroups.com.