Three-dimensional sound reproduction based on binaural synthesis over loudspeakers, including Optimal Source Distribution (OPSODIS), has been extensively studied. In conventional procedures, an inverse filter matrix derived from a measured plant matrix incorporating head-related transfer functions (HRTFs) enables equalization of the reproduction transfer function and crosstalk cancellation, simultaneously. When the playback sound sources contain spatial cues derived from HRTFs, as in binaural recordings, the system can reproduce an immersive and accurate 3-D sound. In contrast, if such information is absent, as in conventional stereo recordings, the perceived image may become excessively widened or result in in-head localization. In this study, the authors propose a discretized source arrangement based on the principle of OPSODIS that emphasizes only crosstalk cancellation without performing transfer function equalization. The method controls only the relative gain and delay between the direct and cross components at each frequency band without system inversion. By avoiding equalization, the approach reduces sensitivity to individual HRTF variation. The proposed method is expected to enhance the lateral spatial impression and perceptual immersion in the absence of explicit HRTF-based cues, provided that appropriate stereo recording conditions—such as setting the microphone spacing to an equivalent interaural distance—are applied.
Yairi et al. (Wed,) studied this question.