ML Based Noise Detection

Electromyographic noise is present in all Electrocargiogram signals because ECG sensors rely on direct contact to the skin. Our alogrithm can classify whether or not an ECG signal has an acceptable amount of noise or is unusably polluted by EMG noise.

Using a novel feature set and a linear regression algorithm from the sklearn toolkit for Python, we developed our classifier that has an accuracy of 91% which we think can be improved with a larger data set. Check out the separation of our classes for each of the six features selected below.

Take a look at our presentation here.