Time-Frequency Signal Synthesis and Its Application in Multimedia Watermark Detection

Abstract

A novel approach is proposed in this thesis to synthesize the time domain chirp signal from the joint time-frequency distribution (TFD). The objective is to reconstruct the original signal from its corrupted version. The new signal synthesis technique is based on the Discrete Polynomial Phase Transform (DPPT) and the TFD of the signal to be synthesized. The TFD is used to separate the mono-component signals from a multi-component signal. The DPPT is then applied on the estimated mono-components to have a final synthesized version of the individual time domain signals. The candidate TFD to be used in the synthesis technique is chosen from a group of common TFDs based on their performance with different types of signals. The criteria for the comparison are joint time-frequency localization, low susceptibility to noise, cross-term suppression and the precision of the instantaneous frequency estimated from these distributions. Smoothed Psuedo Wigner-Ville Distribution is chosen as the processing TDFD in the proposed signal synthesis technique. The proposed chirp synthesis technique is applied to detect the presence of the chirp signal embedded as a watermark message in multimedia security applications. The technique can detect the presence of chirp signals from a corrupted chirp with a bit error rate up to signal synthesis is proved to be less than that of the detection method based on the Hough Radon Transform and the proposed signal synthesis technique may also be used as an error correction tool in other applications.