![denoiser iii error denoiser iii error](https://user-images.githubusercontent.com/1431904/55417867-31b7a200-557a-11e9-974c-eda23874e153.png)
To solve the music noise caused by the location estimation error of effective data and noise distribution, Pascal and Filho and Plapous et al. Singh and Sridharan, Kamath Sunil and Loizou, Sim, Oh and Lee further reduced the music noise in high S/R by improving the SS. reduced the music noise in the case of low and medium S/R. , Mcaulay and Malpass, and Lockwood et al. The SS has been widely used in enhancing speech signal due to its advantages of good denoising function and fast calculation speed. The typical method is spectral subtraction (SS), which is a transform domain filtering method. On this basis, the time-space filtering method was developed and widely used. For example, band-pass filtering, τ- p filtering, and F-K filtering are used and/or combined to remove noise in seismic ahead-prospecting (SAP), tunnel reflection tomography (TRT), and so on, further improving the S/R of data. To solve this problem, time-frequency (TF) information was analysed and introduced in data processing in tunnel environment. Evaluation of noise using frequency domain information cannot meet the requirement of denoising in tunnel environment. However, traditional denoising methods are based on the assumption of stable noise, which cannot be suitable for the nonstationary noise in the tunnel well.
![denoiser iii error denoiser iii error](https://flashbackj.com/wp-content/uploads/2017/06/01143357/Denoiser-Faster-large-980x560.jpg)
In addition to band-pass filtering, τ- p transformation is widely used to separate effectively reflected waves from the waves with different apparent directions and then achieve data denoising. It is necessary to study denoising methods for improving imaging results.
![denoiser iii error denoiser iii error](https://i.redd.it/db0wy9tlena51.png)
In seismic forward-prospecting in the tunnel, the S/R of data is seriously affected by the environmental noise. For example, Sun proposed an adaptive method for seismic data denoising Li proposed an adaptive denoising method based on ensemble empirical mode decomposition (EEMD) and Wang used wavelet transform to eliminate and suppress the random noise. The typical time-domain method consists of convolution filtering, singular value decomposition (SVD), etc., and the transform-domain method consists of F-K filter, Radon transform filter, and time-frequency (TF) filter. The denoising methods can be divided into time-domain method and transform-domain method. To improve the imaging accuracy, the denoising method is developed. However, because of the limited space and construction noise in the tunnel, the seismic data obtained in tunnel environment often contain stronger noise than data collected in ground surface, which affects the extraction of effective signals and leads to anomalies in imaging results. Similarly, seismic data with high S/R are also the premise of seismic forward-prospecting in the tunnel. The data acquisition with good signal-to-noise ratio (S/R) plays an important role in the exploration of oil and gas reservoirs, providing good basis for imaging underground geology. In addition, simulation tests were conducted by using different S/R, which indicates the effectiveness of the TF-SS in removing the random noise. According to the characteristics of effective signal duration between seismic data and voice data, the time-frequency spectrum selection method and filtering coefficient are modified. On this basis, the main frequency information of seismic data is calculated and used to optimize the filtering coefficients. In particular, the proposed method significantly outperforms conventional denoising methods in low signal-to-noise (SNR) conditions.In view of the key problem that a large amount of noise in seismic data can easily induce false anomalies and interpretation errors in seismic exploration, the time-frequency spectrum subtraction (TF-SS) method is adopted into data processing to reduce random noise in seismic data. As a result, the proposed GAN-AE based denoiser (GAD) can effectively can effectively reduce the noise level and has the ability to accurately recover the peak time of the wave packet. The trained denoising AE has strong robustness. Second, according to the characteristics of the guided wave, a pair of datasets are generated to train DAE network. First, GAN is used to estimate the distribution characteristics of the extracted noise and generate samples. To tackle this problem, this paper proposed a denoising network based on the combination of generative adversarial network (GAN) and autoencoder (AE). The traditional denoising method cannot estimate and use the prior information of the signal, which leads to poor denoising effect. In the detection using guided waves, the signal often carries a high level of non-Gaussian noise.