In this thesis work, algorithms are proposed for FSK signal detection and its modulation parameter estimation from digitally modulated unknown signal in voice-band. The modulation parameters of interest are characteristic frequencies and the symbol rate of an FSK signal. Most of voice-band FSK signals are continuous phase modulated to have smaller sidelobes in frequency spectrum and ease corresponding demodulation. The proposed algorithm uses the relative number of occurrences of near-zero envelope to discriminate an FSK signal from the signals which have phase discontinuities between symbols like PSKs or QAMs. The zero-crossing intervals are used to obtain characteristic frequencies of the signal.
The symbol interval is derived from the minimum interval between inter-symbol transitions (IST). ISTs are detected from the abrupt changes of zero-crossing intervals. In order to eliminate adverse effect of noise on the detection of zero-crossing intervals, we propose a new method to detect them. The proposed method builds a signal which has amplitudes of zero-crossing intervals subtracted by their mean value. From the zero-crossings of the signal, we can obtain ISTs. The signal of ISTs is then low-pass filtered to suppress spurious zero-crossings introduced by noise.
Computer simulation is done to obtain the performance of the proposed algorithms. The simulation results show excellent discrimination performance and also show that estimated modulation parameters are precise enough to be used for non-coherent detection.