Prediction of localization error in generating a focused source

Abstract

This paper proposes a method of predicting human localization error for a focused source. A focused source is a virtual source located in between of a loudspeaker array and a listener. However, generation of the focused source cannot avoid the artifact due to causality, listeners always perceive pre-echoes before the desired sound. Since the human hearing system is sensitive to a preceding waves, it can lead a listener to perceive a virtual source in undesired direction. Because the repeating pre-echoes are observed for ineligibly long interval (∼100 ms), it is not clear to distinguish timbral distortions and echoes from the localization error due to the summing localization of the human auditory system. Therefore, a suppression condition was defined from the precedence effect to separate the localization error from timbral distortions. After applying the suppression condition, the energy vector model was used to quantify the localization error. Combining the suppression condition and the energy vector model, localization error in horizontal plane for each listening spot considering positions of focusing point, array shape, driving solutions, spatial sampling, and truncation can be predicted. The examples show that the prediction method clearly holds up with focused source observations reported from relevant literature.