ABSTRACT

All auditory perception is based on sensing sound, and sound is a form of vibration, which arrives at our sensory system as minute changes in the sound pressure impinging on our ear drums. These pressure changes can be faster or slower, and larger or smaller in amplitude. Formally that is all there is to it, even if this very simple mechanistic description does describe the way we perceive sound subjectively rather poorly. For humans, the upper frequency limit of hearing is at about 20,000 Hz. Hence, if we wanted to capture the information in a sound signal digitally, then, according to the Nyquist sampling theorem, we would require at least 40,000 samples per second. Sample rates of around 48 kHz are the current industry standard. Similarly, if our auditory system wants to work with a complete and faithful description of incoming sounds, it will need to have sufcient data bandwidth to process the equivalent of just under 50,000 pressure values per second. To work out the required information rates, we also need to know how wide a range of pressure values needs to be encoded, and the accuracy with which it would need to be encoded.