Synthesis of a wavelet transform using neural network
Jan Stolarek · 2009
Wavelet transform has a wide area of application in signal processing. However there is no single wavelet perfectly suitable for every task. In practice Daubechies 4 is the most commonly used wavelet, since it is well suited for analysis of many natural signals and it oers a straightforward interpretation of the results. It would be very useful to develop a method for adaptive synthesis of a wavelet transform suitable for particular task. Artificial neural networks oer such ability. So far this approach wasn’t explored. This paper presents neural network for synthesis of orthogonal wavelet transform and a method of unsupervised training of this network.