Summary In this paper, 51 sets of directional spectrum measured data collected from the Bohai offshore site were applied to study the comparative characteristics of Wen's direction spectrum and Donelan direction function. Firstly, the optimal frequency value of direction spectrum is analyzed from two aspects: the direction cumulative distribution and the direction distribution function fitting quality. Secondly, combined with the experimental results, the distribution characteristics of different direction functions are compared from the maximum and standard deviation of the distribution function. The experiments prove that Donelan function contains all the states of the wind and wave growth stage, which can fully measure the influence of the wave age on wind frequency spectrum. Then, this paper analyzes the classical Monte Carlo method and the Las Vegas method from the perspective of the computational complexity and direction distribution difference of simulation step. Based on the classical Monte Carlo method, this paper proposes a two‐dimensional random rough sea surface numerical calculation model suitable for different water depths and different wind wave growth stages. Finally, by analyzing the characteristics of the nonlinear sea surface, this paper studies the effects of wind speed, wind direction, and fetch on sea surface wave height and growth state.

Based on the computer simulation technology of zero memory nonlinearity (ZMNL), this paper combines with the backscattering features of sea clutter and conducts simulation for four typical backscattering coefficient empirical models of sea clutter, namely, Technology Service Corporation, Georgia Institutes of Technology, Hybrid Sea Clutter Model, and Naval Research Laboratory. According to the results, the signals of sea clutter simulated by zero memory nonlinearity are able to well satisfy the requirements of spectrum characteristics and amplitude distribution according to experiments on the fitting characteristics of simulated signals. After that, the above-mentioned sea clutter scattering coefficient fitting characteristics are applied to the finite difference time domain (FDTD) electromagnetic scattering equation of the numerical method, and the optimal recursive solution to the backward scattering area fitted by the above four semi-empirical models is proposed for the first time. Then, the method of discretizing the power spectrum equation of the differential form is combined with the FDTD discrete form. The numerical results of discrete simulation take the form of the random phase ω between 0 and 2π and then are converted into a basic iteration of one-dimensional rough length, thus realizing time domain power spectrum inversion. Finally, combined with the above conclusions, this paper proposes a new statistical model based on sea surface backscatter coefficient to invert the two-dimensional sea surface, that is, the three-parameter statistical model. At the same time, this paper further deduces the backscattering function of sea clutter based on random distribution and points out that the incident frequency and the rubbing angle are the main factors affecting the sea clutter scattering model. Based on this conclusion, this paper combines the sea clutter estimation model and the directional function model to simulate a two-dimensional random rough sea surface satisfying the physical laws of the ocean.INDEX TERMS Zero memory nonlinearity, empirical models, finite difference time domain, twodimensional sea surface.

Facial paralysis refers to a facial nerve disordering, with which people may lose the abilities to accurately control their facial muscles for certain facial performances. The diagnosis of such disordering is mainly based on the observation of patient's face in terms of the facial spatial information, such as facial asymmetry. Up to now, this area is still dominated by therapists' subjective examinations clinically. Therefore, automations for this task receive wide attentions in both academic and industrial fields. Recently, the deep learning based methods, the convolutional neural networks (CNNs) more specifically, demonstrate their competitive performance compared with traditional approaches in many areas. However, due to the lack of the structured/labelled facial paralysis data as training data, those deep learning based solutions are still not able to fully attach their superiorities to the facial paralysis evaluation tasks. Another essential aspect for automation in facial paralysis analysis is the facial spatial information extraction. Semantic segmentation is a better choice than traditional template-based facial landmark detection for analysing facial paralysis images, which contain faces in uncommon patterns. However, most existing semantic segmentation approaches are made for indoor or outdoor scene parsing. To this end, this paper presents a deep learning-based approach for automatic facial paralysis grading prediction. The proposed model utilizes a cascaded encoder structure, which explores the advantages of the facial semantic feature for facial spatial information extraction, and then benefits the facial paralysis assessment. A dual-stage cascaded training process is adopted to utilize a mixture of normal and paralysed faces as training data, which exports a well-trained deep neural network model for facial paralysis evaluation. Experiments are conducted in two aspects to demonstrate the performance of each components of the proposed model. Encouraging results are illustrated compared with several existing approaches in the related areas.INDEX TERMS Facial nerve paralysis, dual-stage, cascaded neural network, facial attribute segmentation, paralysis grading prediction.

Based on the linear wave superposition model, we realize the numerical simulation of three-dimensional (3-D) surface waves combined with JONSWAP spectrum and stereo wave observation project (SWOP) directional function. According to the formation characteristics of freak waves to concentrate the wave energy at a specific location, the component waves are modulated. A complete numerical simulation model of time-invariant 3-D freak waves evolution is first proposed in this study. Then, the accuracy of the model is verified from the aspects of wave height distribution, frequency spectrum estimation, and freak wave parameters. The effectiveness of wave steepness as the discrimination condition of freak waves is discussed through experiments. In terms of the electromagnetic scattering characteristics of freak waves, we construct an electromagnetic scattering model, fitting the time-invariant 3-D freak wave, based on the two-scale method (TSM). By comparing and analysing the scattering characteristics D-value of synthetic aperture radar (SAR) image of the freak wave and the background wave, the rationality of the electromagnetic scattering characteristics of the freak wave as its feature identification is verified. Comparing the normalized radar cross section (NRCS) of freak waves and background sea waves, the experiment shows that the NRCS value of freak waves is the lowest, and the calculation results of the two have obvious differences. The research conclusions above can provide effective data support for the identification and detection of freak waves in practical offshore engineering.

Based on fifty one groups of data on direction distribution measured from buoy sites, several important spectrum parameters including distribution characteristics of the measured data’s spectrum, the Wen’s direction spectrum and the Donelan function are analyzed from the perspectives of standard deviation of directional distribution function and statistical results. Then, a numeric calculation model based on the Monte Carlo method is proposed in this work. At the same time—based on Weierstrass self-affine fractal function—numeric simulation of random sea surface is conducted by modifying the bilateral power law. The analysis of the numeric calculation results under different wind directions, speeds and fetches proves that both methods can be adopted for different water directional distributions and target spectrum models. In addition, this study compares the characteristic wave within different distribution frequency domains in the main wave direction and in its orthogonal direction. It is proved that the fractal method cannot fully reflect the anisotropy of gravity wave and tension wave in the superposition direction, however, it can maintain the similarity of overall energy part with the rough length of the waves. Moreover, there are still limitations of the fractal method in the numeric modeling of undeveloped sea surface.