In this thesis, we investigate some statistical feature–based algorithms for automatic recognition of digital modulation signals. In order to judge the type of a modulated signal, we compute some discriminating statistics of the instantaneous amplitude, frequency and phase of the digital modulation signal. There are three steps for the research. First, we generate the different test digital modulation signals and analyze their features. Next, we use the artificial neural network for “automatic recognition of digital modulation signals”. Last, we propose to exploit the probability density function (PDF) of amplitude, frequency and phase to extract more distinguish statistic of the signal’s feature. Comparing with the traditional neural network method, it is more reliable and easy to extend to more different case. In addition, it has other merits like low complexity and superior performance in low SNR environment.