A machine learning project to identify heart sounds and classify them as an anomaly or a normal cardiac sound, S1 or S2.
We used some audio files containing various recordings of heartbeat in different conditions. Then, we developed various machine learning models based on the perceptron neural networks. The models were trained using cut audio files, that only contained a fraction of the total audio, with either an S1 or S2. The models were then applied to new uncut audio files, where we calculated the heartbeat rate and obtained the classification at specified times.
- Machine Learning applied to audio analysis
- What are heart sounds?
- Data used to train and test
- Other approaches
Beatriz Negromonte (negromontebs@gmail.com)
Francisco Relvão (franciscofrelvao@gmail.com)
Oriented by Filipe Veloso (filipe.veloso@cern.ch)