Complexity In Signal Processing Using Cepstral Approach

373, which is better as with another method suitable for the real-time cepstral processing, i.e. the adaptive phase integration, at the same complexity level. 1 Introduction In many applications, signals that are overlapped in time and have approximately the same frequency contents have to be separated. It means the problem of nonlinear filtering which, however, is transformed to a simple linear filter if the theory of generalized superposition [1] is applied. Generalized superposition is based on homomorphic systems [2] consisting of an input characteristic system that converts a nonlinear operation to a linear one, e.g. convolution to addition, a linear system, and an output inverse characteristic system. Denote by x(n), h(n), and y(n) a system input, unit-sample response, and output signal, respectively. Let y(n) be a convolution<F