Wavelet Based Signal Processing

Project Leader: Dr Brian Ng (UoA)
Project Team: Si Tran Nguyen (UoA)

Funding: University of Adelaide, Australian Postgraduate Award

Project background

Wavelets have long been established as an effective time-frequency analytical tool for processing many classes of real-world signals, with successes in applications ranging from pulse signal denoising to image compression. Traditional discrete wavelet transforms (DWT) are based on a series of  multirate filterbanks using sampling rate changes with dyadic factors. A drawback of this approach is the one octave step changes in the tradeoff between time and frequency localizations at the various subbands as the depth of the DWT increases. Analysis of chirp-like signals, which are commonly found in radar systems, may benefit from finer step changes in time/frequency localizations. This can be achieved in a DWT framework by using rational sampling rate changes in place of dyadic factors, although such a change will require completely new wavelet filters to be designed. This project aims to design novel wavelet filters which will fit into the DWT framework and permit their use in existing applications. Of particular relevance to radar signal processing is wavelet-based denoising. Our new designs will be simulated against current approaches to quantify any improvements to radar signal processing based on metrics such signal-to-noise ratios and time-of-arrival estimates.

Project aims

• Design novel wavelet transforms
• Develop software modules for the processing of radar signals
• Use simulations to verify the effectiveness of the novel algorithms and compare the results with existing techniques
• Seek to cooperate with DSTO to acquire real data