Previous studies showed that the distortion of the spectral information by occluding the pinna cavities of one of the ears decreases the accuracy of vertical localization. However, it is unclear whether vertical localization can be accomplished using the spectral information at only a single ear or requires the spectral information at both ears. In the present study, median plane localization tests in which the head-related transfer functions (HRTFs) for different target vertical angles were presented to each ear of the subjects were carried out. The results showed that the subjects either localized a single sound image to the target vertical angle presented to either the left or right ear or localized two sound images to both target vertical angles, with approximately the same accuracy as when identical target vertical angles of HRTFs were presented to both ears. These results imply that the spectral information at only a single ear provides the spectral cue for vertical localization.
We have developed a sound-selective hearing support system by making use of an environment sensor network, so that individual target and anti-target sound sources in the environment can be selected, and spatial information of the target sound sources can be reconstructed. The performance of the selective sound separation module was evaluated under different noise conditions. Results showed that signal-to-noise ratios of around 15 dB could be achieved by the proposed system for a 65 dB babble noise plus directional music noise condition. Subjective intelligibility tests were conducted in the same noise condition. For words with high familiarity, intelligibility scores increased from 67% to 90% for normal hearing subjects and from 50% to 70% for elderly subjects, when the proposed system was applied.
Boundary surface control (BoSC) is a useful method of reproducing the sound field physically. However, it is challenging in the case of real-time operation. This is mainly due to the calculation cost of a lot of inverse filter convolutions to obtain reproduction signals. This paper proposes a method for reducing of number of inverse filters and implements it in 24-channel narrow directivity (shotgun) microphone array and 24-channel circularly arranged loudspeaker array. Moreover, it provides an experimental evaluation of the reproduction accuracy according to measurement of reproduced wavefront. The accuracy of the reproduced wavefront by the filters, whose number was reduced to less than 1/5 by proposed method, was comparable with the case of full number of filters. Finally, a system aiming at sound field reproduction in a wide frequency range was constructed by a hybrid method of reproducing with an inverse filter in the low range and directly outputting from the speaker in the direction corresponding to the microphone in the high frequency range. We confirmed that real-time processing is possible for this hybrid method by using a convolution plug-in of digital audio workstation software.