In audio processing you deal with spikes using one or two averaging filters with different time constants for rise/fall.

You can also do it with a PID approach