On Feature based Automatic Classification of Single and Multitone Signals

A.K. Das, P. Arabshahi, T. Wen, and W. Su (USA)


Blind demodulation, signal classification, multitone signals.


We consider the problem of feature based automatic classification of single and multitone signals. Our objective is to extend existing blind demodulation techniques to multitone waveforms such as MIL-STD-188-110B (Appendix B) and OFDM, developing a capability to identify signal types based on short data records, and maintaining robustness to channel effects. In this paper, we report on the first phase of our approach, namely, building a coarse classifier for a range of single tone and multitone signals. Among the features considered by the coarse classifier are those based on trigonometric moments and higher order statistics of the instantaneous frequencies of the received signal. No a priori information is assumed on the part of the received signal. The received signal of interest has not been previously observed; it is not part of a library of known signals; and no automated classifier has been built for it. Extensive simulation results based on real world signals are presented demonstrating the feasibility of the above features for automatic classification purposes of single and multitone signals.

Important Links:

Go Back