Blind deconvolution via cumulant extrema——基于累积量的盲反卷积算法.pdf

lassical deconvolution is concerned with the task of variable, and the corresponding moments and cumulants of recovering an excitation signal, given the response of that random variable as found in standard probability text- ’ a known time-invariant linear operator to that excita- books (e.g., [13,16,17,26]). These concepts are then applied tion. In this article, deconvolution is discussed along with its to the study of stationary random time series and the task of more challenging counterpart, blind deconvolution, where no deconvolution. knowledge of the linear operator is assumed. This discussion In a typical signal processing application, one is concerned focuses on a class of deconvolution algorithms based on with the task of extracting information contained in a set of higher-order statistics, and more particularly, cumulants. experimentally obtained data as designated by {x(n)}.It is These algorithms offer the potential of superior performance useful to interpret the elements of this data set as being in both the noise free and noisy data cases relative to that samples of a sequence of underlying random variables as achieved by other deconvolution techniques. This article denoted by {X(n)}.The indexing variable, n, exclusively provides a tutorial description as well as presenting new takes on integer values and is frequently associated with time, results on many of the fundamental higher-order concepts although other descriptors may be more appropriate in a used in deconvolution, with the emphasis on maximizing the given application (e.g., distance, temperature). Thus, a dis- deconvolved signal’s no zlizc random variable s