To solve the problems of color distortion and structure blurring in images acquired by sensors during bad weather, an image dehazing algorithm based on feature learning is put forward to improve the quality of sensor images. First, we extracted the multiscale structure features of the haze images by sparse coding and the various haze-related color features simultaneously. Then, the generative adversarial network (GAN) was used for sample training to explore the mapping relationship between different features and the scene transmission. Finally, the final haze-free image was obtained according to the degradation model. Experimental results show that the method has obvious advantages in its detail recovery and color retention. In addition, it effectively improves the quality of sensor images.

The Doppler effect causes frequency offset and spectrum spread of communication signals, which increases the difficulty of feature extraction when identifying the modulation mode in aeronautical wireless channels. To improve the signal modulation recognition capability of the single monitoring node in Wireless Sensor Networks (WSN), a method based on the bidirectional convolution gated recurrent deep neural network (Bi-CGDNN) with an attention mechanism is proposed in this paper. We design the framework with different types of network layers, realizing frequency offset suppression and feature mapping. In addition, quadrature sampling signals and instantaneous feature statistics are collected to generate the dataset. Thus, the waveform features and phase-frequency information of the signals can be used for combinatorial analysis. In particular, the attention mechanism is introduced to optimize the framework for better feature perception and higher recognition rates. Experimental results show that our method achieves great recognition effects at different maximum Doppler frequencies. The recognition rate reaches 98.1% at the signal-to-noise ratio (SNRs) of 8dB with the Doppler frequency of 200Hz.

Current modulation recognition methods in wireless sensor networks rely too much on simulation datasets. Its practical application effect cannot reach the expected results. To address this issue, in this paper we collect a large amount of real-world wireless signal data based on the software radio device USRP 2920. We then propose a real radio frequency (RF) database architecture and preprocessing operators to manage real-world wireless signal data, conduct signal preprocessing, and export the dataset. Based on different feature datasets derived from the RF database, we propose a multidimensional feature hybrid network (MFHN), which is used to identify unknown signals by analyzing different kinds of signal features. Further, we improve MFHN and design a multifeatured joint migration network (MJMN) to identify small-sample targets. The experimental results show that the recognition rates for unknown target signals of the MFHN and MJMN are 82.7% and 93.2%, respectively. The proposed methods improve the recognition performance in the single node of wireless sensor networks in complex electromagnetic environments, which provides reference for subsequent decision fusion.

When communicating in aeronautical wireless channels, the difficulty of radio modulation recognition increases due to the loss of information caused by noise; particularly in circumstances with low signal-to-noise ratios (SNRs), it is difficult to achieve recognition rates exceeding 90.0%. To improve the radio modulation recognition performances of networks at low SNRs in complex electromagnetic environments, a modulation recognition method based on multidimensional feature analysis is proposed in this paper. It is realized through a cascaded structure including a Deep Cross Network (DCN) and an improved Visual Geometry Group Network 16 (VGG16). Our network framework is divided into two modules. In the one-dimensional data analysis module, we take the high-order cumulant of a transmitted signal as the one-dimensional feature input of the DCN. In the two-dimensional data analysis module, the color constellation density of the signal is extracted as the feature map input of the improved VGG16. Finally, we build a cascaded neural network with hybrid feature inputs for modulation recognition. Experimental results show that the recognition rate of our method is higher than 90.0% at an SNR of -4 dB. Compared with other methods, the proposed method has better recognition performance at low SNRs in aeronautical wireless channels.

Massive user generated social media texts (SMT) posits new opportunities as well as challenges for psycholinguistic analysis to understand individual differences such as personality. SMT are often written in an informal way, and thus contain lexical variants such as nonstandard spellings, capitalizations, and abbreviations. These lexical variants are referred as out-of-vocabulary (OOV) words. They are not captured in standard dictionaries used by standard Natural Language Processing (NLP) tools. Literature indicates that these OOV words may include hidden linguistic patterns that reflect individual characteristics. These OOV-related linguistic patterns are not captured by the existing closed-vocabulary and open-vocabulary approaches. To address these issues, this dissertation develops two artifacts, following a design science research process model. The first artifact is an OOV-aware data curation process that focuses on capturing and categorizing OOV words. The evaluation of the first artifact demonstrates that it can capture more OOV words and is useful in analyzing SMT. The second artifact is an OOV-aware hybrid approach that integrates the closed-vocabulary and open-vocabulary approaches with expanded OOV categories and OOV words. The hybrid approach shows an improved performance over existing approach. This dissertation makes theoretical contributions by adding additional OOV knowledge and a new method for psycholinguistic analysis of SMT. It also makes practical contributions by enabling psycholinguistic researchers and practitioners to exploit more psycholinguistic cues for tasks like personality prediction.