Continuous Wavelet Transform in Real Time : A block-based approach

Sammanfattning

Continuous wavelet transforms (CWT) are a powerful tool for analyzing signals with non-stationary frequency characteristics, but their computational complexity has prevented widespread adoption in real-time applications. Recent advances in the field have produced the fast continuous wavelet transform (fCWT), which claims to accelerate computation by at least an order of magnitude without loss of accuracy. This thesis investigates and evaluates the fCWT, as well as builds on their work to investigate whether it makes real-time applications of the CWT computationally feasible. It uses the reference implementation of the fCWT provided by the authors, and further develops it to perform real-time transforms of streaming data. Some flaws in the fCWT algorithm are identified, and their consequences are analyzed. The results show that the fCWT does provide a sufficiently large speedup of computation to enable real-time applications.