Unless you are hanging out in a cave with lots of bats, noise is not too big of a problem. (Bats actively use ultrasonic sound in the 20-80 kHz range for echolocation). Using frequencies in the ultrasonic range is very nice, as most things in the world do not emit sounds with these frequencies. In a standard home or office environment, frequencies above about 25 kHz are relatively quiet. Therefore, noise is not a big consideration.
One group (Bass & Bolen) measured ultrasonic noise associated with various manufacturing operations with an eye on where ultrasonically controlled robots can be used. They found that industrial operations such as impact, bending, grinding and drilling provide little ultrasonic energy above 100 kHz. For these processes, ultrasonic sensors operating above 100 kHz should have an advantage since signal and noise will both fall off with increasing frequency. However, they found that ultrasonic emission from laser etching was found to be quite broadband, making such processes less suitable for ultrasonic control. High-velocity fluid, or air sprays, are the most prevalent sources of ultrasonic noise. The noise from these sources rolls off slowly with frequency. Basically, in frequency bands below several hundred kHz, there is some noise, but it is usually not appreciable. Above a few hundred kHz, things get very quiet, and noise is usually negligible.
One uninvestigated consideration is noise from electronic equipment, such as computers and computer monitors. It has been suggested that these kinds of devices might give off appreciable sound in the 20 kHz to 100 kHz. If given more time, I would investigate this area more fully, but since so many people have told me that it will not be a problem, I will assume that to be so.
Internal System Noise
A bigger noise concern is eliminating noise from the system itself. "Glitches" or spikes can be internally generated by the electronics used for the communications system. The best way to deal with this is to carefully choose your transducers (next section). Echoes from the signals will produce a kind of noise as well. The best way to deal with echoes is with a sturdy data protocol (see Vadim's work in acoustic FSK and AM).