Phillip L. Ainsleigh
Naval Res. Lab., Underwater Sound Reference Detachment, Orlando, FL 32856-8337
Charles K. Chui
Texas A & M Univ., College Station, TX 77843-3368
A technique is introduced for removing unwanted components from data using B-spline-based multiresolution (B-wavelet) analysis. The time and frequency localization capabilities of wavelets allow two advantages over standard filtering methods: (1) better noise detection arising from the ability to isolate noisy components in wavelet bands where the signal energy is small relative to the noise energy, and (2) less signal loss by selectively filtering only those segments within wavelet bands containing unwanted noise. The procedure is used to remove impulsive noise from time series, and to eliminate multipath interference from frequency domain data.