Page 13 - 《应用声学》2019年第6期
P. 13

第 38 卷 第 6 期               张国昌等: 使用同心多环阵提升声源定位鲁棒性                                           915


                 与仿真结果类似, 在 CH 展开阶数相同时,                          [5] Salvati D, Drioli C, Foresti G L. A low complexity robust
             UCCA 的定位鲁棒性在所有实验布放的声源位                                beamforming using diagonal unloading for acoustic source
                                                                   localization[J]. IEEE/ACM Transactions on Audio Speech
             置下均具有明显优势。分析CH展开阶数,在大部分
                                                                   & Language Processing, 2018, 26(3): 609–622.
             的声源方位下,UCA 在 CH 展开阶数为 4 阶时定位                        [6] 居太亮. 基于麦克风阵列的声源定位算法研究 [D]. 成都: 电
             性能较优,但与 3阶CH 展开的 UCCA 相比,仍有不                          子科技大学, 2006.
                                                                 [7] 柯昆. 声源定位技术研究 [D]. 西安: 西安电子科技大学,
             小的差距。
                                                                   2010.
                 综合上述的仿真和实验结果来看,UCCA 在                           [8] Brandstein M, Ward D. Microphone arrays: signal pro-
             以下条件下均具有显著优于 UCA 的定位性能:                               cessing techniques and applications[M]. Springer Science
                                                                   & Business Media, 2013.
             (1) 相同CH展开阶数,上述所有的T 60 和SNR条件                       [9] Lee B, Kalker T. A vectorized method for computationally
             下;(2) 强混响或者低信噪比时,任意的 CH 展开阶                           efficient SRP-PHAT sound source localization[C]. 12th In-
             数下。在高噪声和弱混响的部分 CH 展开阶数下,                              ternational Workshop on Acoustic Echo and Noise Con-
                                                                   trol, F, 2010.
             UCA与UCCA具有接近的DOA 估计性能。                             [10] Zhang C, Florencio D, Ba D E, et al. Maximum likeli-
                                                                   hood sound source localization and beamforming for direc-
             5 结论                                                  tional microphone arrays in distributed meetings[J]. IEEE
                                                                   Transactions on Multimedia, 2008, 10(3): 538–548.
                 本文阐述了一种针对多环阵列的环谐波展开                            [11] Ishi C T, Chatot O, Ishiguro H, et al.  Evaluation of
                                                                   a MUSIC-based real-time sound localization of multi-
             技术。环谐波展开系数经补偿滤波器补偿后被用于
                                                                   ple sound sources in real noisy environments[C]. 2009
             声源方位估计。为了提升CH域方位估计的稳健性,                               IEEE/RSJ International Conference on Intelligent Robots
             针对 UCA 贝塞尔函数零点处噪声放大问题,提出                              and Systems, F, 2009.
                                                                [12] Cobos M, Marti A, Lopez J J. A modified SRP-PHAT
             了一种最小模准则设计 UCCA 补偿滤波器的方法。                             functional for robust real-time sound source localization
             针对混响环境,利用环谐波展开解耦频率与方位角                                with scalable spatial sampling[J]. IEEE Signal Processing
             相关性的特点,CSSM 被用于UCA 和UCCA。仿真                           Letters, 2011, 18(1): 71–74.
                                                                [13] Silverman H F, Yu Y, Sachar J M, et al. Performance of
             结果表明,合理设置孔径的最小模 UCCA 可以解决                             real-time source-location estimators for a large-aperture
             UCA 零点问题,同时也避免了正则化引入的低频                               microphone array[J]. IEEE Transactions on Speech & Au-
             波束展宽。统计性能的仿真和实验显示,在相同麦                                dio Processing, 2005, 13(4): 593–606.
                                                                [14] Traa J, Wingate D, Stein N, et al. Robust source lo-
             克风数目和阵列孔径的前提下,与 UCA 相比,使用                             calization and enhancement with a probabilistic steered

             UCCA可以显著提升混响环境下的DOA稳健性。                               response power model[J]. IEEE Transactions on Speech &
                                                                   Audio Processing, 2016, 24(3): 493–503.
                                                                [15] Zhang C, Florencio D, Zhang Z. Why does PHAT work
                                                                   well in low noise, reverberative environments?[C]. IEEE
                            参 考     文   献
                                                                   International Conference on Acoustics, Speech and Signal
                                                                   Processing, F, 2008.
              [1] Evers C, Naylor P A. Acoustic SLAM[J]. IEEE/ACM  [16] Wang H, Kaveh M. Coherent signal-subspace processing
                 Transactions on Audio Speech and Language Processing,  for the detection and estimation of angles of arrival of mul-
                 2018, 26(9): 1484–1498.                           tiple wide-band sources[J]. IEEE Transactions on Acous-
              [2] Nikunen J, Virtanen T. Direction of arrival based spa-  tics Speech and Signal Processing, 1985, 33(4): 823–831.
                 tial covariance model for blind sound source separation[J].  [17] Doron M A, Weiss A J. On focusing matrices for wide-
                 IEEE/ACM Transactions on Audio Speech and Language  band array processing[J]. IEEE Transactions on Signal
                 Processing, 2014, 22(3): 727–739.                 Processing, 1992, 40(6): 1295–1302.
              [3] Farmani M, Pedersen M S, Tan Z H, et al. Maximum like-  [18] Yoon Y S, Kaplan L M, McClellan J H. TOPS: new DOA
                 lihood approach to “informed” sound source localization  estimator for wideband signals[J]. IEEE Transactions on
                 for hearing aid applications[C]. 2015 IEEE International  Signal Processing, 2006, 54(6): 1977–1989.
                 Conference on Acoustics, Speech and Signal Processing.  [19] Claudio E D D, Parisi R. WAVES: weighted average of
                 New York, IEEE, 2015: 16–20.                      signal subspaces for robust wideband direction finding[J].
              [4] Minotto V P, Jung C R, Lee B. Simultaneous-speaker  IEEE Transactions on Signal Processing, 2001, 49(10):
                 voice activity detection and localization using mid-fusion  2179–2191.
                 of SVM and HMMs[J]. IEEE Transactions on Multimedia,  [20] Tianaroig E, Jacobsen F, Grande E F. Beamforming with
                 2014, 16(4): 1032–1044.                           a circular microphone array for localization of environ-
   8   9   10   11   12   13   14   15   16   17   18