Statistics of input signals for frequency domain identification of weakly nonlinear systems in communications

Transmission systems in communications give rise to linear and nonlinear distortions. Therefore, system identification procedures applied to verify their performance should not only supply the "best fit" to the linear frequency response but give also an estimate of the nonlinear distortion spectra. However, the nonlinear distortions strongly depend on time and frequency domain statistics of the input signal. The main idea of the paper is to design the test signal so that it reproduces the statistical behaviour of typical input signals. Only then the identification will yield relevant results for certain applications. At the same time the advantages of periodic test signals for FFT processing are to be retained. The estimation procedure for frequency domain identification of weakly nonlinear systems is described and an outline on the influence and design of test signal statistics with applications for testing of audio coders in communication technology is given.