Localization performance of a two-listener virtual sound imaging system

This study describes the design and evaluation of a crosstalk cancellation (CTC) system intended to reproduce binaural audio for two listeners simultaneously. A compact loudspeaker array was combined with measured loudspeaker-to-ear transfer functions to generate band limited virtual sources and frequency domain CTC filters. The signal processing framework, including the implementation of the crossover network and the construction of the frequency-dependent CTC matrices, is presented in detail. The resulting filters are examined in both the time and frequency domains to illustrate their behaviour and to show how individual loudspeakers contribute across frequency bands. While the theoretical foundations of the array geometry are established in a companion paper, the present work focuses on the physical implementation of the system and its perceptual evaluation through subjective listening tests in which 25 participants localized virtual sound sources reproduced at two listening positions. After correction for front-back confusions, a mean absolute error of approximately 15° was achieved at both listening positions, with no statistically significant overall difference in accuracy or consistency between the two seats. The results indicate that listeners were generally able to identify the intended source directions, demonstrating that stable spatial cues can be delivered simultaneously at both positions.