Filtering of Periodic Noise Using the Complex Wavelet Transform

Claus Benjaminsen

AbstractEngines, compressors and other machinery performing cyclic processes produce a special kind of noise, which can be called periodic noise. This very common phenomenon - often loud - can create great difficulties, when trying to communicate verbally with another person. With the signal processing possibilities in cell phones and other telecommunication devices, this disturbance can be removed.

In this report a periodic noise filtering scheme is presented based on nearly analytic complex wavelet packets with good shift invariant properties. The shift invariance comes from the Dual-Tree Complex Wavelet Transform, which the nearly analytic complex wavelet packets are built on. But in order to fully maintain the good shift invariant properties of the Dual-Tree Complex Wavelet Transform, the extension to wavelet packets can not be done straight forwardly. It turns out that a special ordering of the wavelet packet filters is needed, and that specific ordering giving nearly analytic complex wavelet packets is developed and presented in this report.

The developed periodic noise filtering scheme gives promising results compared to a spectral subtraction scheme in both a measure of the signal to noise ratio and in a subjective listening test. The scheme calls for some further improvements and tests, but has a potential of making its way into tomorrows
telecommunication devices.
TypeMaster's thesis [Academic thesis]
PublisherInformatics and Mathematical Modelling, Technical University of Denmark, DTU
AddressRichard Petersens Plads, Building 321, DK-2800 Kgs. Lyngby
NoteSupervised by Assoc. Prof. Jan Larsen, IMM, DTU.
Electronic version(s)[pdf]
BibTeX data [bibtex]
IMM Group(s)Intelligent Signal Processing

Back  ::  IMM Publications