7.4 HRTF set dependencies

7.4.1 Experimental procedure

In this experiment, HRTFs measured for different subjects are compared to investigate potential effects of various HRTF sets. Four different HRTF sets of the CIPIC HRTF database were evaluated in this test. These are the sets ‘018’, ‘137’ and ‘147’ as well as the set ‘003’ that was already used in the previous experiment. The comparison of these HRTF sets was done for positions 1 and 9. Position 1 was selected because it resulted in a low w threshold in the previous experiment and it represents a very common sound source position. Position 9 was selected since it is positioned more to the side and has a slightly different elevation component.

The experimental procedure, the subjects and the stimuli were identical to the procedure employed in the previous test. Only one parameterization method (using overlapping parameter bands) was used to determine the thresholds for w.

7.4.2 Results and discussion

The results are shown in Figure 7.5. The two tested positions for each of the four HRTF sets are marked on the abscissa. The corresponding thresholds of the warp factor w averaged across three trials of the five subjects is given along the ordinate. The errorbars denote the standard deviation of all measurements.

The results for HRTF set ‘003’ are quite similar to the data obtained from the previous experiment; position 1 has a threshold value of about 1.4, while position 9 results in a threshold of about 2.1. For the other three HRTF sets, however, thresholds are generally lower for corresponding positions. A minimum of 1.0 is obtained for HRTF set ‘137’, position 1. The tendency for HRTF sets ‘003’, ‘018’ and ‘137’ that position 9 gives rise to higher thresholds is not observed for HRTF set ‘147’.

Figure 7.5 Thresholds for the warp factor w for the four CIPIC HRTF sets (‘003’, ‘018’ ‘137’ ‘147’) and two positions (‘1’ and ‘9’).

The mean thresholds averaged across subjects found in this experiment range from 1.0 to 2.1, a range that is still in line with our hypothesis that thresholds are expected to be at least 1.0. In this respect, position 1 of HRTF set ‘137’ showed to be most critical.

The results also indicate that different HRTF sets may give rise to different threshold values, and that directional dependencies may differ across HRTF sets. In other words, there does not seem to be one single position or sound source direction that can be characterized by a low number of parameter bands in a general sense.