Magnetotelluric signal denoising method and system based on time-frequency domain fusion network

• Main Authors: LIANG QING, CHENG HONG, ZHAO XIAOLIN, YE JINGKUN, LI NIANCHUN, LUO YUCHENG, LI JIN, WANG JIAYU
• Format: Patent
• Language: Chinese; English
• Published: 23.07.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS

The invention discloses a magnetotelluric signal denoising method and system based on a time-frequency domain fusion network, and the method comprises the steps: constructing a noisy signal sample library and a noise contour sample library or a pure signal sample library; training a noise reduction mapping model of a fast Fourier transform residual network based on a bidirectional coordinate attention mechanism based on the magnetotelluric signal sample and a noise contour sample thereof or a pure signal sample library; then, the magnetotelluric signal to be subjected to noise reduction is input into the noise reduction mapping model to obtain a pure signal or a noise contour, the noise contour is subtracted from the magnetotelluric electric field signal to be subjected to noise reduction to obtain the pure signal, and the magnetotelluric signal is subjected to noise reduction. According to the technical scheme of the invention, the fast Fourier transform network based on the bidirectional coordinate attentio