Combined likelihood power estimation and multiple hypothesis modulation classification
K.M. Chugg, Chu-Sieng Long, A. Polydoros · 2002
Previously developed techniques for maximum likelihood (ML) modulation classification have assumed that there are only two possible modulation formats and that both the signal and noise powers are known. We introduce ML-based techniques for performing autonomous power estimation of a phase-shift-keyed signal and additive white Gaussian noise, and for classifying between OQPSK, BPSK and QPSK formats. The performance of the ML power estimator is shown to be superior to existing techniques and the false classification rate of the simple, two-stage OQPSK/BPSK/QPSK classification rule is shown to be close to that of the globally optimal classifier. A fully autonomous QPSK/BPSK/QPSK classifier is demonstrated by combining the two-stage rule, the ML power estimator, and previously developed threshold-setting techniques.