The Cambridge loudness model: Design principles and predictions

Abstract

The Cambridge loudness model, which forms the basis for the new ANSI standard (ANSI S3.4-2005), is similar in structure to loudness models proposed earlier by Fletcher and by Zwicker, but differs in many of the details. The input signal to each ear is specified in terms of its spectrum, which may be defined in 1/3-octave bands, or as the sum of a series of sinusoidal components and one or more pink-or white-noise bands. The first stage of the model is a filter to account for the transfer of sound from the sound field or headphone to the eardrum. The second stage is a filter to account for the transfer of sound through the middle ear. The third stage transforms the effective spectrum reaching the cochlea to an excitation pattern, calculated as the output of an array of level-dependent auditory filters. In the final stage, the excitation at each center frequency is converted to specific loudness, by a nonlinear frequency-dependent transformation. The specific loudness is summed across frequency to give the loudness for each ear, and is summed across ears to give overall loudness. The model accounts well for the effects of level, frequency, and bandwidth on loudness.