Multiple input linear models are commonly used for noise source identification. For stationary noise signals, conditioned spectra may be computed and contributions of the output coherent to each input may be found. In the general case the inputs are partially coherent, which may lead to errors when interpreting the contribution of each input to each output. A method is presented which separates the output spectrum into components that are coherent only to each input as well as jointly coherent to two or three inputs. The proposed method is verified by a computer simulation with two partially coherent inputs. Steps to apply this method to nearfield acoustical holography with two or three reference signals are presented. (C) 2008 Institute of Noise Control Engineering.
展开▼