利用压缩网格方法提高旋转声源定位计算效率

Improving the computational efficiency of rotating sound source localization via compression computational grid method

利用两种压缩网格方法:基于传统波束形成的CG2压缩网格方法和基于小样本数据的CG4压缩网格方法,对经典时域旋转声源波束形成(ROSI)算法进行优化加速。实验结果表明:两种压缩网格方法均不影响ROSI算法旋转声源定位效果,基于传统波束压缩形成的CG2网格方法能够提高ROSI算法的旋转声源定位计算效率1~2倍,基于小样本数据的CG4压缩网格方法能够提高ROSI算法的旋转声源定位计算效率13~18倍。除此之外,基于小样本数据的CG4压缩网格方法在相控麦克风阵列平面与旋转声源平面垂直条件下,仍能准确进行旋转声源定位。

In order to improve the computational efficiency of the classical time-domain rotating source identifier(ROSI) beamforming, two different compression computational grid methods were proposed: one based on conventional beamforming, namely CG2, and the other one obtained from ROSI beamforming with a little amount of sample data, namely CG4. Experimental applications showed that both compression grid methods did not affect the effectiveness of the ROSI algorithm for rotating sound source localization. ROSI beamforming with CG2 can improve the computational efficiency of the ROSI algorithm for rotating sound source localization by a factor of 1 to 2, while ROSI beamforming with CG4 can improve the computational efficiency of the ROSI algorithm for rotating sound source localization by a factor of 13 to 18. In addition, ROSI beamforming with CG4 can still accurately locate the rotating sound source even if the microphone array plane run perpendicular to the rotating sound source plane.