| Blind methods for communication-system signal processing have become essential and widespread, due in part to their enabling fast synchronization, higher information rates, multipoint transmission, and non-cooperative reception. Yet, modern communications signals of all types increasingly do not have one of the most essential and standard properties required for blind statistical processing: independent and identically-distributed symbols. Undesirable source statistics may be produced for a variety of intentional and unintentional reasons, including constellation shaping, pseudorandom source sequences, source coding, and nonlinear modulation. The development of new blind methods suitable for use with such sources would further increase the utility and flexibility of these important signal processing techniques.; Many blind signal processing techniques were designed for constant-modulus signals but are applicable to or can be extended to multi-modulus signals (the best known example being the constant modulus algorithm); some blind methods make direct use of modulus to process multi-modulus signals. Constant-modulus signals were and remain popular in situations where significant amplitude distortion is possible, including mobile radio and satellite communications. Continuous-phase modulated signals, in particular, are popular because they can be relatively bandwidth efficient, they can achieve low out-of-band power, and simple receiver designs can be used. Such schemes have recently found significant use in wireless communication environments, notably in the GSM, Tetrapol, and HiperLAN standards. Multi-modulus signals are more complicated to transmit and receive, and they are more susceptible to channel distortion, but they can have high spectral efficiency.; We develop new methods for blind channel estimation and equalization for constant-modulus and multi-modulus signals, focusing primarily on methods that accommodate or exploit temporal correlation. We apply existing constant-modulus-type methods to multi-modulus signals, we describe new methods for blind channel estimation for continuous-phase modulation, and we develop new modulus-exploiting methods of blind equalization for multi-modulus signals. In particular, we develop several new blind equalization methods that exploit known modulus, and we use these in new turbo-driven methods for blind equalization of, possibly temporally-correlated, multi-modulus signals. Compared with existing blind and non-blind turbo methods, the new turbo-driven methods easily accommodate fractionally-spaced equalizers, they can potentially be implemented in a fully blind manner, and they can accommodate larger constellations and longer channels. |
