
时延估计算法地方法很多.docx
4页峠值佈计检测结果时延估计算法的方法很多,广乂互相关函数法(Gee, Genear I i zedeross-ocerrat Inin)运用最为广泛广 义互相关法通过求两信号之间的互功率谱,并在频域内给予一定的加权,来抑制噪声和反射的影响,再 反变换到时域,得到两信号之间的互相关函数〃其峰值位置,即两信号之间的相对吋延45IH, 6],时延估 计过程如图1一 7所示”设h. (n), h (n)分别为声源信号s (n)到两麦克风的冲激响应,則麦克风接收到的信号为:2Xi (n) =hi (n) 0S (n) +ni (n) (1. 1)x (n) =h (n) 0 s (n) +n (n) (1.2)2 2 2X1X202y 1 (n) 畑心1) *~►峰值 检测 结OF * * I“疾 y2(n)X2(n) ——► y 2(n) IH<@)基于子空间的定位技术来源于现代高分辨率谱估计技术子空间技术是阵列信号 处理技术 中研究最多、应用最广、最基本也是最重要的技术之一该类声源定位 技术是利用接收信 号相关矩阵的空间谱,求解麦克风间的相关矩阵来确定方向角,从而进一步确定声源位 置子空间类方法主要分两类,一类是利用阵列自相关矩 阵主特征向量(即信号子空间)的 主分量方法,如AR参数模型主分量法,BT主 分量法等;另一类方法是以信号子空间和噪 声子空间的正交性原理为基础,利用 组成噪声子空间的特征向量来进行谱估计,这类算法 主要有多重信号分类法(MUSIC), Johnson 法,最小范数(Mini-Norm)法,MUSIC 根(Roo t-M USIC)法,旋转不变信号参 数估计(ESPRIT)法,等等。
在实际中,基于子空间的定位技术 的空间谱的相关矩阵是未知 的,必须从观测信号中来估计,需要在一定时间间隔 内把所有信号平均来得到,同时要求 接收信号处于声源、噪声、估计参数固定不 变的环境和有足够多的信号平均值即便满足 这此条件,该算法也不如传统的波 束形成方法对声源和麦克风模型误差的鲁棒性好目前 定位问题所涉及算法都是 研究远场的线性阵列情况基于子空间的定位技术是通过时间平均来估计信号之间的相关矩阵,需要信号是 平稳过 程,估计参数固定不变,而语音信号是一个短时平稳过程,往往不能满足 这个条件该定 位方法的效果和稳定性不如可控波束形成法,但每次迭代的计算 复杂度不像可控波束形成 那么苛刻此外还要求阵列处于远场情况,而且该方法 主要针对窄带信号,而语音信号是宽带信号虽然如此,在适当的改进后,也可 用于近场环境[51] 王永良,陈辉,彭应宁等,空间谱估计理论与算法,北京,清华大学出版 社,2004年门月[52] Kr im H. , Viberg M,Two decades of array signal processing research, I EEE SiganProcess ing MgaZa i ne, 1996,13(4),[53] Kay S. M. , Marple S. L. , Spectrum ana lysis, Proc, of the I EEE, 1981,11.[54] BugrJ P.,Maximum entropy spectral ana lysis, Proc.of the 37」meeting of the Annua I I nt.SEG Meet i ng, Ok IahomaC i ty, OK, 1967[52] Krim H., Viberg MMTWO decades of array signal processing research, IEEE Sigan Processing Magazine, 1996,13(4), page: €7-94.[53] Ka-y S,M.» Marple S.L„ Spectrum analysis-a modem perspective. Prg of the IEEE, 1981» 11.[54] Buirg J.R, Maximum entropy spectral analysis. Proc, of the 37th meeting of the Annual lot. SEG Meeting, Oklahoma City,OK, 1967[55] Capon J. High-resolution frequency-wavenumber spectrum analysis. Proc, of IEEE, 1969, 57(8):1408 〜1418 ・[$6] Kumaresan R., Tufts D・W“ Estimating the angels of aniyal of plane waves. IEEE Frans. OnAES. 1983, 19(1): 134-139[57] Schmidit R.O. Multiple emitter locateon and sigaal parameter eslitnattion, IEEE bans., 1986, APt34(3): 274280.[58] Cadzow J.A., Kim Y.S., Shie D.C., General diiection of arrival estimateion: a signal subspace approach, IEEE Tran J Ort AES, 1989, 25(1): 31—46.刃 Rao B.D ・》Kari K.V.S., Performance analysis of Roo 卜 MUSIC, IEEE Trans. On ASSP, 1989, 37(12): 1939-1949.[60] Kung S.Y,, Arun K.S., Rao State space and SVT) based approximation methodsfor the harmonic retrieval proplem, J. Opt. Soc. Amer., 1983,73(12): 1799-1811 ・[61J Rao R・ ,Kailath T・* ESPRIT-a subspace rotation approach to estimation of parameters of cissoids in noise, IEEE Trans. On ASSP, 1986, 34(10): 1340-1342.[62] Rao R・ ,Kailath T・ ,ESPRJT-estimatian of signal parameters via rotational invariantcc techniques, IEEE Trans, on ASSP, 1989,37(7): 9&4T93.f63] Stoica P・ ,Nehorai A・ ,MUSIC. Maximum likelihood, and Cramer-Rao bound, In Proc. ICASSPl 1988,22942299 ・[64J Ottetsfen D・ ,Viberg. M., Stoica P., Nehorai A.,Exact and large sample ML techniques for parameter estimateion and detection in array processings In Haykin, Litva, and shepherd, editors. Radar array processing, Springer-Verlag, Berlin, 1993.99-151.[65] Cadzow J.A. A higji resolution direction of arrival algorithm for narrow-band coherent andtincoherent sources, I EEE Transen ASSP. 1988, 36(7): 965 〜979.[66] Cler&eot H., Tressens S., Ouamri A., Perfennance of high resolution frequencies estimateion methodscompared to the Cramer-Rao bound? IEEE Trans, on ASSR【989. 37(11): 1703-1720.3[67j Ziskind I., Wax M. Maximum likelihood localization of multipJe sources, by alternating projection, IEEE. Trans. OnASSR 19&8(6(10): 1553-1539.[6&] Stoica R, Sharman K.C., Novel eigenanalysis method for direction estimateion, IEEE・ Procewling, Pt .F, 19^0,137(1): 19~26.[69] Viberg M., Ottersren Kailath T., Detection and estimateion in sensor arrays using wighwd subsp^e fitt ing, IEEE, Tbs, on SP, 1991, 39(11): 24342499.[70] Bresler Y・,Macovski A. Exact maximum likelihood parameter estimation of superimposi-d5exponential signals in noise, IEEE. Trans. On ASSP,. 19X6, 34(5): 108HI089.[16]S・ Doclo, Multi-microphone noise reduction and dereverberation techniques for speech applications, Ph.D. Thesis, ESAT, Katholieke University Leuven, Belgium, Chapter 9,Near-FieldBroadband Beamforming, pp217-232,2003.Panayiotis q Georgiou, Chris Kyriakakis, An Alternative Model for Sound Signals Encountered inReverberant Environments; Robust Maximum Likelihood Localization and Parameter Estimation Based on a Sub-Gaussian Model, Audio Engineering Socicty(AES) 113th Convention Paper, 2002, Los Angeles,USA。
