Cyclostationarity based blind classification and parameter estimation for FSK and MSK signals

Blind signal classification and parameter estimation plays an important role in both military and civilian applications. The classification and estimation task provides signal information, such as modulation type, carrier frequency, signal bandwidth, and symbol timing, for the design of effective communication systems. In general, blind signal classification and parameter estimation is very challenging, particularly in environments involving a low signal-to-noise ratio (SNR) regime, short observation periods, fading channel conditions and relaxed a priori information. -- Due to its easy implementation and widespread usage in legacy communications equipment, the frequency shift keying (FSK) modulation continues to be very common, especially in the VHF and UHF bands. On the other hand, minimum shift-keying (MSK) scheme is also widely used in wireless communication systems as it possesses many advantages, such as bandwidth efficiency and constant-envelope property. Thus, the blind classification and parameter estimation of FSK and MSK signals becomes an attractive research area. Most of existing approaches for FSK and MSK signal classification and parameter estimation require pre-processing such as symbol timing and carrier recovery, and only additive Gaussian noise (AWGN) channel is considered. -- In this thesis, the cyclostationarity-based FSK and MSK signal classification and parameter estimation are studied. The first- and second-order cyclostationarity of FSK and MSK signals affected by fading is investigated. Based on the first-order cyclostationarity of FSK signals, a joint classification and tone frequency spacing estimation algorithm is proposed. Furthermore, a symbol period estimation algorithm for FSK signals is proposed based on the properties of second-order cyclostationarity. By combining the properties of first- and second-order cyclostationarity of FSK and MSK signals, a joint classification algorithm for FSK and MSK signals is proposed. Simulation and experimental results are canied out to show the efficiency of proposed algorithms. It is proved that reasonably good performance can be obtained at low SNRs, using short observation period, under the fading effect, and with relaxed a priori information.