There are frequencies of sound that fall outside of the range of human hearing, high frequencies are still sound pressure, and do the tiniest of waves, and slow low frequencies that are so slow that it is more like a wind than a thump.
If I was trying to make water move with sound pressure, but to not hear the same sound, I would try and make some kind of very slow bass alogrythm that averaged out the same sound , then moved the water "to the sound" still.
I would be talking about a theory, if I had not already seen it in action, and played with it some.
Remember those "activators" for computers and consoles that a person could strap on thier chest , or mount in a chair. The activator would convert electrical energy from the sound to more physical movements. It did this with a huge weighted magnet assembly , passing through a electromagnetic coil.
In these devices they had a switch that would half or quarter the sound waves, taking the same sound and making movements that are 1/2 or 1/4 the frequency. They basically translated all the midrange frequencies to lower, so the physical action of the activator was greater for the mid-low frequencies too.
One of the issues with the physical activators, no mater how much of the sound was being converted to physical movement and not sound output, sound output occurs. everything that is physically moved can create a sound pressure wave. (is why the "voice coil" method of armature arm in hard drives still makes some sound)
At any of the lower freqencies to move the speaker (or thse activator electromagnets) these large distances takes much electrical energy, speaker or activator. By lowering the frequency the ammount of physical movement is longer the power greater. (low frequencies want to move the magnetic stuff longer distances to create the low sounds, more work = more power)
So one way to make water dance to music without hearing as much sound output from the electromagetic activator would be to lower the frequency of the sound it was getting. Many ways to do that, like to change the playback/sample rate without resampling. Extreeme pitch adjustments, octive change and "time" changes all ways to lower the frequency.
Problem still is you would want to lower the freqency only to the activator and not the speaker too, or your going to need an unadjusted speaker output to mask the noises. So any alogrythm would want to be a curcuit in the device, and might also need more power going to the water activator.
That still leaves the best solution what tech1337 provided, to stop or reduce the output to the speakers designed for audible output, and still maintain power to the water activators.