Multiple Beamforming with CSP-Based Source Localization
Takanobu Nishiura, Satoshi Nakamura, Kiyohiro Shikano · Institutional Repositories DataBase (IRDB) · 2000
In real environments, the presence of room reverberations seriously degrades sound capture quality. To solve this problem, multiple beamforming, which forms directivity not only in the direction of the target sound source but also in the direction of reflection images, has been proposed by J. Flanagan, et al. However, it is difficult to apply the method practically, since it requires an accurate source position and room geometry. This paper proposes a new multiple beamforming algorithm which requires no prior information about the source position or room geometry. The proposed algorithm achieves multiple beamforming by estimating the position of the target sound source and the major reflection images by CSP (Cross-power Spectrum Phase) analysis. Evaluation experiments based on the image method are conducted. The proposed algorithm performs more effectively than the conventional single beamformer in a room with large refiection coefficients. However, the proposed algorithm performs almost the same as conventional single beamforming in a room with small reflection energy.