Noise effect reduction obtained by using coherence function in deconvolution

The behavior of a linear system may be completely described by its impulse response, step response or frequency response. They can be theoretically calculated from time records of the input and output of the system. It is a deconvolution problem and is ill posed in particular owing to the noise present in both the measured input and output. This paper briefly describes three deconvolution methods based on optimization techniques and compares their results with those obtained by a non iterative criterion based on coherence function. An experimental application of the proposed method is also presented.