A Dictionary-Based Pursuit Algorithm for Magnetotelluric Signal-Noise Separation

A critical challenge in magnetotelluric (MT) studies is the effective suppression of noise in collected data prior to investigating deep geological structures and detecting deep-seated blind ore bodies. To address this issue, this study proposes a novel method for magnetotelluric signal-noise separation based on stagewise orthogonal matching pursuit. The approach involves constructing a redundant dictionary comprising wavelet packets, cosine atoms, and other basis atoms to sparsely represent segmented MT data. A Gaussian matrix is then employed to observe and sample the sparsely represented data. Subsequently, the stagewise orthogonal matching pursuit algorithm is applied to suppress strong interference and accurately reconstruct useful MT signals. The method is validated using both simulated typical strong interference and measured MT data from an ore concentration area. Comparisons with matching pursuit, orthogonal matching pursuit, and the remote reference method demonstrate that the proposed technique offers higher computational efficiency, fewer iterations, better convergence, and more effective noise suppression. Furthermore, the calculated apparent resistivity-phase curves exhibit improved continuity and smoothness, and the quality of low-frequency MT data is significantly enhanced.