AbstractSince most of the noise generated by electronic equipment using high‐frequency devices has statistical properties that are quite different from Gaussian noise, the bit error rate of conventional receivers is greatly degraded in a man‐made noise environment, because they are designed with the presumption of Gaussian noise. One of the methods of providing a better bit error rate in such a noise environment is the use of an optimum receiver. This optimum receiver attains a high bit error rate by using maximum likelihood symbol detection based on the statistical properties of the noise, and the statistical properties of the noise must be known a priori for optimum reception. In this paper, we use the Middleton class‐A impulsive radio noise model as the statistical model of the noise and study the parameter estimation and bit error rate of an optimum receiver in a class‐A impulsive radio noise environment. First, we evaluate the effect of errors in parameter estimation on the bit error rate of the optimum receiver in a class‐A impulsive radio noise environment, and show that to improve the performance of the optimum receiver in this noise environment, we should use parameters that express the statistics of the noise. Next we study the relationship between the number of noise samples used in estimation and the parameter estimation accuracy, and demonstrate that large numbers of noise samples are needed for highly accurate parameter estimation. However, it is shown that a quite good bit error rate can be obtained by the optimum receiver in the noise environment even if parameter estimation is based on a somewhat smaller number of noise samples (for example, about 700 samples for an impulsive index A = 0.05 and a Gaussian component ratio Γ′ = 0.05). © 2003 Wiley Periodicals, Inc. Electron Comm Jpn Pt 1, 86(8): 68–78, 2003; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecja.1173