Rectification and low-pass filtering

One aspect that is commonly found in measurement models such as this is a simulation of the breakdown of phase-locking in the neural output of the ear above a certain frequency. Whilst complex models of this process have been developed, it is commonly mimicked by the use of half-wave rectification and low-pass filtering. Bernstein and Trahiotis investigated the effect of half-wave rectification and low-pass filtering on the use of the IACC in predicting the detection of binaural differences in a range of stimuli. They found that the use of half-wave rectification and a low-pass filter prior to the IACC calculation closely matched their empirical results [Bernstein and Trahiotis 1996].

This was investigated as part of the project, and it was found that the inclusion of imitation of the breakdown of phase locking in the ear is necessary to accurately predict the perceived width of higher frequency stimuli [Mason et al 2005a]. This causes the fine temporal detail of higher frequency stimuli to be lost, with the perceived width being dependent on the cross-correlation of the envelope of the signal.

The processing in the model includes passing the input signal through half-wave rectification and a 6th order Butterworth low-pass filter prior to the cross-correlation calculation. This was found to match the results of subjective judgements most accurately.