Data Processing Methods for Time Frequency Peak Filtering

Abstract

Time-frequency peak filtering is a valid method for attenuation of random noise in the bandlimited nonstationary deterministic signals. In order to reduce the zero-drift errors resulted from processing practical data by TFPF, the fixed zero scaling method is used here to implement TFPF. The zero-drift error generated in the two procedures of TFPF, frequency modulation and peak filtering in time-frequency plane, is related to signal scaling and length of discrete Fourier transform. The quantized error is in inverse ratio to length of discrete Fourier transform, and the zero-drift error is rooted in the varying value scaled from zero. The fixed zero signal scaling is convert the zero to the fixed value in span of instantaneous frequency to decrease the error resulting from discretization and to improve uneven two dimensional signal by TFPF. The segment TFPF technique is used to ensure the amplitude is even in one segment, which can reduce the discretization error in filtered signal and recover the weak signal. The results of simulation and processing seismic common shot data show that the improved TFPF can attenuate the strong random noise and have the less zero-drift errors. This improvement further to make the TFPF practicability.