Localization in the Presence of a Distracter and Reverberation in the Frontal Horizontal Plane: II. Model Algorithms

Previously, it was shown that humans are able to localize a broadband noise burst quite well, even if it is presented in the presence of a second distracting broadband noise burst at 0-dB target-to-distracter ratio. In this investigation, different model algorithms are tested for their ability to predict the psychoacoustic results. First, a long-time and a running cross-correlation algorithm are introduced to simulate the psychoacoustic experiments. Both algorithms, which are used to analyze the time window in which target and distracter are presented, are not able to estimate the target position as accurately as human listeners. Therefore, a listening test was conducted to find out, if the listeners used the part of the distracter that preceded the target in the previous tests as a reference. The outcome indicates that the onsets of test sound and distracter must be separated temporally; otherwise, human localization performance deteriorates rapidly. These results are incorporated into the interaural cross-correlation difference (ICCD) model, the third model approach that is introduced. For each frequency band, the interaural time differences of the test sound are calculated from the difference in the interaural cross correlation (ICC) of the total sound (target signal+distracter) and the ICC of the distracter alone. The latter is calculated from the part of the distracter that precedes the test sound. The model is able to demonstrate a number of psychoacoustic effects, including shifts in the auditory events for low target-to-distracter ratios.