KEYWORDS: Interference (communication), Control systems, Education and training, Neural networks, Signal attenuation, Denoising, Data modeling, Tunable filters, Head, Signal processing
In active noise control system, the noise control effect based on the traditional method is reduced under nonlinear interference from the electronic devices like speakers and microphones. In this paper, a combined CNN-LSTM network active noise control system is proposed, and an experimental scenario for imitating a real auditory system is built using an artificial head. The primary and secondary channel models are estimated using BP neural networks to simulate the nonlinear properties from the electronic devices. Sever type noise data from various scenes are chosen for simulation. As compared with the conventional FxLMS algorithm, FNN, CNN and LSTM, the trials demonstrate that the proposed network can effectively suppress the noise at both low and high frequencies with high convergence speed.
Ground Penetrating Radar (GPR) is a widely used technology for non-destructive near-ground detection. It is often difficult to balance detection depth and resolution when using a single-frequency antenna. To solve this problem, a novel multi-frequency data fusion algorithm is proposed. The two-dimensional Gaussian weighted window is utilized as a sliding window to analyze regional energy characteristics. By matching multi-frequency data with these properties, the benefits of high resolution for high-frequency data and high amplitude for low-frequency data can be leveraged. The matching threshold is determined using a genetic algorithm. Fusion is then performed based on the matching results obtained. The effectiveness of the proposed algorithm was evaluated through experimental comparisons with other algorithms at frequencies of 900MHz and 1600MHz. The experiment results indicate that the proposed algorithm can enhance the resolution and information content, from which a superior radar profile is acquired.
Buried object detection methods based on deep learning require a lot of annotated data, and most of them rely on pretrained models. To solve these problems, a buried object detection method that only needs a small amount of annotated data and has a short training time is proposed. This method integrates the attention mechanism into the U-net model, obtains the pixel-to-pixel predicted grayscale, and finally extracts the region of interest for target localization. The experimental results show that this method can accurately detect buried objects with only a small amount of annotated data in the actual B-scan images of ground penetrating radar.
Most of the traditional compressed sensing ground penetrating radar imaging is performed in a two-dimensional form, and it is impossible to accurately determine the specific location of the underground target. In order to solve these problems, a three-dimensional imaging method is proposed for ground penetrating radar. In this method, the data is observed from a planar MIMO array, and a dictionary matrix is constructed into a three-dimensional model. Finally, the underground scene in three dimensions is reconstructed with the total variation minimization. The experimental results show that in the layered media scene, the proposed method can accurately image the target by using only a small amount of data.
Deep learning method has been extensively applied to ground penetrating radar two-dimensional profile (GPR B-SCAN) hyperbola detection recently. We propose a B-SCAN image feature extraction method based on the constraints of the GPR physical model, and further detect the weak boundary feature curve of the target in the local space. A deep convolutional neural network (DCNN) is first designed to extract high-level semantic features from B-SCAN images to remove direct wave. Next, a multiscale feature fusion DCNN is used to extract the features of the B-SCAN image with the direct wave removed, and the classifier network is used to identify the hyperbola of the upper boundary feature of the target. Finally, according to the hyperbola, the local space corresponding to the target in the B-SCAN image is determined. On this basis, the amplitude and phase information of the scattered electric field are used to segment the lower boundary characteristic curve of the target through convolution operation. Experimental results on simulation and field data show that feature information of the buried target in the GPR B-SCAN image can be efficiently extracted when the proposed method is adopted.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.