How do I classify heart sounds using Euclidean distance?

I have a set of training signals, (all have murmurs) and a testing signal. How can I use Euclidean distance to identify if it have a murmur or if the heart sound is normal?

Answer 1

[Image Analyst]

At the very least you'd have to crop off any partial beat signals at the ends of the array, then scale the signal to account for different heart rates, which may even require a different scaling for the "between beats" signal and the "during beats" signal. But, I'm very, very certain this has been done before. People have been working on ecg signals for decades. What do published papers say on this? What algorithm do they use? Anything I might tell you would be overly simplistic compared to what's been published, or anyone else unless they happen to be working on ecg signals themselves.

Answer 2

[Shashank Prasanna]

You haven't provided much detail about the nature of the problem, but it appears that kNN classifier might be what you are looking for. kNN uses Euclidean or any other distance metric you specify to classify your data if you already know the response for your training data:

Take a look at the examples at this page:

http://www.mathworks.com/help/stats/classificationknnclass.html

If your training data does not include information about the response, then you do clustering using euclidean distance with KMEANS:

http://www.mathworks.com/help/stats/kmeans.html