I am creating a neural network using sound as input, I need several algorithms to make the sound file manageable.

i need to rectify the signal, then use some kind of envelope/filter to reduce the complexity and vastly reduce the inputs, then finally I need to detect a starting point in the file to remove the unnecessary silennce.  I was thinking that anything below a certain dB would be chopped off from the beginning.

Does anyone know where I can find these,, or find out info on these so I can write them myself?