Search

This paper proposes a channel response generative adversarial network (CRGAN)-based turbo code interleaver that estimates a channel response and interleaver indices at a transmitter by using a sound speed profile (SSP) and the ocean environments without feedback from a receiver. The interleaver indices are designed to allocate important bits from the turbo code to subcarriers with great channel gains, which reduces them from being affected by deep fading. Computer simulations and practical ocean experiments demonstrate that the proposed method estimates the channel response with low mean squared errors (MSEs) and improves bit error rate (BER) performances compared with the conventional method.

This paper proposes a deep learning-based Direction-of-arrival (DOA) estimation to detect interfered far-field sources. The proposed method consists of a near-field interference rejection network (NFIRnet) and a DOA estimation network (DOAnet). The NFIRnet calculates the near-field components of the covariance matrix by convolutional neural networks with the proposed complex mapper. The near-field components are rejected from the covariance matrix. The DOAnet removes the residuals of the interferences by the proposed self-spatial attention network and estimates the DOAs of the interfered far-field sources. Computer simulations demonstrated that the proposed method had better DOA estimation performance than the conventional methods.

In underwater sensor networks (USN), channel impulse response (CIR) estimation based on sound speed profile (SSP) ensures link reliability. We propose a separate bi-directional feature pyramid network (S-BiFPN) that estimates the CIR using deep learning based on SSP. The proposed method enlarges the small variation of SSP by converting 1-dimension into 2-dimension, extracts various features using a fused feature map obtained from the separate paths, and estimates the CIR. Simulations are performed using the practically measured SSPs and CIRs, and the results show that the proposed method has the lowest mean square error (MSE) and a reasonable running time compared to conventional methods.

In underwater acoustic (UWA) sensor network, the channel impulse response (CIR) at the transmitter is important to increase the link reliability and the throughput. The CIR feedback to the transmitter decreases the throughput due to the feedback propagation delay, and the estimation of the CIR at the transmitter is also difficult since the sound of speed profile (SSP) may not be continuously measured. This paper proposes a deep learning based CIR estimator that estimates the SSP from only one water temperature sensor at a depth of the transmitter. The proposed CIR estimator consists of a 2-dimensional temperature network, 2-dimensional bidirectional long short term memory (2D BiLSTM), and a fully connected layer. The proposed algorithm learns the SSP variation with the depth and the time using 2D BiLSTM and estimates the CIR from the SSP. The estimated CIRs are utilized for the multi-user diversity to increase the link reliability and the throughput of the UWA sensor network. The computer simulations and practical ocean experiments were executed to evaluate the estimation error, the bit error rate and the throughput of the proposed algorithm. The proposed CIR estimator demonstrated better performance than other 1D conventional algorithms.

In this paper, we propose a high resolution grid-based deep learning source localization that precisely estimates the locations of near-field sources without a priori knowledge of the number of sources. The proposed method consists of a principal component analysis network (PCAnet) and a spatial spectrum network (Sp2net). The proposed PCAnet calculates the noise spaces of the received signals by convolutional layers without a priori knowledge or the estimation of the number of sources and has the lower complexity than eigenvalue decomposition (EVD). The proposed Sp2net calculates the spatial spectrum with a linear layer from the output of the PCAnet and classifies dense location grids with a convolutional neural network (CNN). From the spatial spectrum, this paper also proposes an activation function to enlarge the values at the grid points where the near-field sources exist, which are differentiable for all input values. Then, the direction of arrivals (DOAs) and the ranges of the near-field sources are estimated with high resolution. Computer simulations demonstrated that the proposed method had better DOA and range estimation performances than those of the conventional methods.

Purpose We aimed to identify the factors associated with the repeated febrile seizures (RFS), defined as recurrent seizures during the same febrile illness. Methods We reviewed the medical records of children with febrile seizure who visited 4 academic emergency departments from October 2016 through September 2018. Differences were identified in variables regarding clinical and laboratory characteristics between the children with and without RFS. The RFS was the primary outcome. Logistic regression was conducted to identify factors associated with the occurrence of RFS. Results Among 1,551 children, 922 were included in the study, of whom, 198 (21.5%) underwent RFS. Of the children with RFS, 188 (94.9%) underwent the recurrences within the initial 24 hours. Logistic regression showed focal seizure (adjusted odds ratio, 6.67; 95% confidence interval, 2.37-18.82), venous pH < 7.31 (5.89; 3.13-11.08), and postictal drowsiness > 30 minutes (1.90; 1.30-2.78) as the factors for RFS. Conclusion In children with febrile seizure, focal seizure, acidosis, and prolonged postictal state may be independent risk factors for RFS. These findings may be informed to healthcare professionals and parents caring for children with febrile seizure.