Signal detection and modulation classification using a goodness of fit test

Cognitive radio increases its importance as a system to realize an efficient use of radio spectrum. In cognitive radio, signal detection and modulation classification are key issues to recognize the radio channel condition. In this paper, new signal detection and modulation classification methods based on distribution tests are proposed. Both the Anderson-Darling test and the Cramer-Von-Mises test are examined for a signal detection error probability and a modulation classification error probability. The modified Anderson-Darling test is also proposed to simplify and speed up the test. Numerical results of error probability show that the proposed methods are effective to identify signals in the radio channel, especially in a case where the received signal has a low signal-to-noise ratio without prior knowledge of signals.