A New Adaptive Filtering Technology In Signal Processing

Adaptive filter theory and applications is one of the important developing aspect of the modern control theory, and it has important theoretical and applied value in nonlinear system identification, modeling, prediction and filtering. In this paper we make a study of the theoretical means and its applications in the domain of oil field well logging and well test.Firstly we have detailedly studied the time-domain, frequency-domain and transform-domain LMS adaptive digital filters. The structure and relative learning algorithm of filters are introduced. Furthermore a lot of simulation and application examples have been done.In fact there are a lot of systems that are polluted by time-variant correlative noise and the convergent speed of LMS algorithms becomes very slow. So, this paper proposes an adaptive noise canceler (ANC) which consists of an adaptive Kalrnan filter and MAP estimator of noise statistics. Simulation results are obtained by using a conventional smooth signal with sinusoidal.components as well as a "non-smooth" signal recorded oil well colour spectrum data with time-variant correlative noise. It has been shown from simulation results that the new ANC is feasible and effective.In order to heighten the precision of signal processing, in the third section a wavelet-based adaptive deconvolution filter is presented by hybridizing the least mean square filter, the Kalrnan filter and the wavelet transform. A lot of simulated experiments have been done by using the scheme. The results show that the filter can track the reflection coefficient signals and cancel the high frequency noises, and the average relative errors is within 1%. This indicates that the scheme is very efficient.