Global Navigation Satellite System Reflectometry (GNSS-R), as a novel retrieval technique, can be effectively used for the retrieval of ocean wind speed. However, existing wind speed retrieval methods have not fully solved the problems of inaccurate reference values and sparse samples in high wind speed areas, resulting in insufficient accuracy in high wind speed retrieval. To tackle this issue, this paper proposes a multimodal wind speed retrieval network based on ResAtt-Net, which further improves the retrieval accuracy in high wind speed areas by combining fine-tuning strategies and wind speed fusion strategies. Experiments based on FY-3E GNOS-II level 1 data at different wind speed ranges show that our model exhibits superiority in both overall and high wind speed retrieval, especially in the high wind speed regime (>17 m/s) where the retrieval error is significantly reduced. The RMSE decreases from 4.32 m/s to 2.65 m/s, and the correlation coefficient increases to 0.897. This method provides an efficient and accurate solution for ocean high wind speed remote sensing retrieval tasks based on FY-3E data, and has broad application prospects.

An ANN model-based method is proposed to retrieve the ocean surface wind speed from Cyclone Global Navigation Satellite System (CYGNSS) L1 observation data. The proposed method is embedded with an input features and output results filtering module based on a machine learning (ML) algorithm, abbreviated as the MF-ANN method. By studying the relationship between physical observation variables from CYGNSS L1 data and the wind speed of the European Centre for Medium-Range Weather Forecast (ECMWF) reanalysis dataset and Cross-Calibrated Multiplatform (CCMP) dataset, a feature parameter selection module based on the XGBoost algorithm and Pearson correlation coefficient is designed to automatically select the optimal combination of feature parameters related to the ocean surface wind speed and input the feature parameters into the ANN model. An outlier filtering module based on the unsupervised ML algorithm is designed to identify and eliminate the outliers in the wind speed data retrieved by the ANN model, reducing the errors that cannot be eliminated by the fixed threshold of the quality flag providing in the CYGNSS L1 product. The experiments show that the RMSEs of the MF-ANN model can be reduced by 11.73% and 15.31% compared with those of a single ANN model by adding feature selection modules and quality filtering modules, respectively. The RMSE of the MF-ANN model combining these two modules can be reduced by 29.08%.

In recent decades, indoor positioning techniques have been researched to support automatic guidance for visitors in public buildings such as museums, galleries, etc. The exhibition goods nearby could be introduced by a smartphone after the location of the visitor has been acquired. This paper presents an indoor positioning scheme utilizing smartphones equipped with inertial and magnetic sensors. The positioning method could handle complicated human motion and various ways to hold the phone. This is achieved by applying an improved Pedestrian Dead Reckoning (PDR) algorithm and an error-tolerant magnetic map matching algorithm. The improved PDR algorithm estimates walking distances and direction in real time. As long as the user stops to appreciate showpieces, the improved PDR component will report relevant data to magnetic calibration component. Based on the updated information, the final location of the user could be calculated utilizing the relevant data, the real-time data of magnetic field sensor and the magnetic map acquired in advance. We evaluate this method using experimental measurements in practice.

This paper presents two novel low-complexity encoding algorithms for quasi-cyclic (QC) codes based on Galois Fourier transform. The key idea behind them is making use of the block diagonal structure of the transformed generator matrix. The first one, named encoding by Galois Fourier transform, is equivalent to the fast implementations of the traditional encoding by Galois Fourier transform. The second one, named encoding in the transform domain (ETD), requires much less computational complexity for encoding binary QC codes. It skips the first step of the first algorithm and applies post-processing to save a large number of Galois field multiplications. Its application to QC-LDPC codes is also studied in this paper. Particularly, the hardware cost of the ETD for RS-based LDPC codes can be greatly reduced by short linear-feedback shift registers.