The Research On The Method Of Wavelet Transform Based On The The Switching Current Circuit

The wavelet transform is a new way for signal analysis, which named fordecomposing the original signal to a linear combination of the different scale of thewavelet. The most popular way for wavelet transform is software method and digitalcircuit method. Although, these method is easy to achieve, but they have poorefficiency, specially for the low power application,such as cochlear implant andembedded system.So analog method for Wavelet Transform is tredding to be the mostpopular way for the low power application of wavelet transform.Switch current circuit is new technique of analog signal sampling,which promisesto provide the benefit of low power, high-precision and easiness to be compatible withdigital process, etc.The core technology of analog method is approximation of function, which is howto change the function of wavelet to a rational fraction effectively. The most commonway is Pade approching.But Pade method just have effect in a small region nearing theexpansion point. The farther away form the expansion point, the lower Signal to NoiseRatio it has. Method of time demain approching is easier than frequemcy demainapproaching, but there are few people study about it.In this context, we focus attention on study a new way for time demainapproching. The main content of this dissertation are listed as follows:1.With the technology equivalent interval sampling and using power function forprimary function, we keep information of pole-zero to the time domain. So we can usethe sampled data to get the pole-zero easily.2.Use mathematics way to change the wavelet function to a ARMA(p,p) model.So it is easy to use principal component analysis to get order of system and get thecoefficient of system with linear least square method.3.Because of the original system is non-linear, using a linear system to appoarch awavelet function might leading many problem. So we introduce iteration after the firstparameter estimation. This iteration use the parameter of first estimation for the startpoint of Levenberg-Marquardt method.