Multiresolution transient detection

Designs and studies the performance of a multiresolution-based transient detector. The transients the authors are interested in consist of wide-band, pulse-like, coherent structures in a turbulent flow. To take advantage of the fast pyramidal wavelet algorithm, an important point when processing large amounts of experimental data, the detector makes use of the discrete wavelet transform. The authors show how the lack of time-invariance drawback of the discrete transform can be efficiently overcome by using relevant analytic wavelets. They thus compare this detection technique with one based on a continuous wavelet transform, as well as with other standard methods and show that wavelets perform best when the transients are superimposed on a colored 1/f background noise. This description is very close to that of turbulence and relevant also in many other situations. >