extract features from raw vibration signals using sparse filtering

I am trying to implement a way to use sparse filtering or another similiar algorithm to extract features directly from raw vibration signals. I tried the algorithm:

     sparsefilt

But the results were not great. Maybe because this algortihm doesnt like to take raw signals as input (but wants already features?)

Also I have to mention that I am aware of this method:

https://de.mathworks.com/help/predmaint/examples/Rolling-Element-Bearing-Fault-Diagnosis.html

But my task is to find a way to exract features more automatic.

Can somebody help me to find the right algorithm to extract features from raw vibration signals?

Best regards

Answer 1

[Wang Pan]

Same problem met.

Maybe Try PCA or ICA methold.