Frequency and loudness compensation

Necessity of compensation for frequency and loudness

Previous studies have indicated that the relationship between the measured IACC of a sound and the perceived width is dependent on the frequency of the sound [Blauert and Lindemann 1986, Yanagawa et al 1990]. In addition, previous studies have indicated that the relationship between the measured IACC of a sound and the perceived width is dependent on the loudness or level of the sound [Morimoto and Iida 1995].

These additional dependencies in the relationship between the IACC and width need to be taken into account in order to create a measurement that is as widely applicable as possible. If this is not taken into account, measurement results will only be directly comparable if the frequency and loudness of the sounds is identical. In order to make directly comparable measurements of a wider range of sounds, some form of compensation must be employed to take into account the differences in frequency and loudness.

Frequency dependence of the relationship between IACC and width

We undertook a detailed investigation into the causes of the frequency dependent relationship between the measured IACC of a sound and its perceived width [Mason et al 2005a]. Three main factors were found to have an effect on this. Firstly, it was found that it is necessary to simulate the breakdown of phase-locking in the ear, as discussed previously. Secondly, it was found that there is a strong dependence on frequency at low frequencies, possibly caused by the limited range of IACC values created in natural sound fields at these frequencies. Thirdly, it was found that even for stimuli that were identical at both ears, the width was dependent on frequency.

Methods were derived to compensate for these three factors, as described in [Mason et al 2005a], and these are applied to the measurement model as described below.

Level or loudness dependence of the relationship between IACC and width

We undertook a detailed investigation into the loudness dependent relationship between the measured IACC of a sound and its perceived width [Mason and Brookes ????]. The results of this experiment showed that this dependence could be modelled by multiplying the measured IACC by a single loudness dependent factor that was independent of the IACC value or the frequency of the stimulus.

Application of these factors to the model

As it appears from this information that both loudness and frequency have an effect on the perceived width of a stimulus, this needed to be included in the measurement model to increase the accuracy and applicability of the results. The frequency dependence is modelled in the measurement using three processes, one in advance of the cross-correlation calculation and the remainder in this segment of the model. Firstly, the input signal is rectified and low-pass filtered as described previously. Secondly, the measured result is multiplied by a frequency-dependent factor that is based on the expected range of IACC values at low frequencies. Finally, a model of the frequency dependence of the perceived width when the IACC = 1 is added to the result.

For the level dependence, the loudness of the input signal at a given point in time as derived previously is used to determine the loudness dependent factor by which the results are multiplied.

The resulting values from these processes are then forwarded for further analysis.