Desert seismic signals contain strong random noise，which brings great trouble to the processing and interpretation of desert seismic signals． In order to solve this technical problem，Deep Ｒesidual Convolutional Neural Network for Shearlet Transform model is proposed for the implementation of the desert seismic signal random noise suppression． In training phase，the Shearlet coefficients of desert seismic data are taken as inputs，and the Shearlet coefficients of random noise are taken as labels． Through network training，the mapping relationship between them could be learned by a deep CNN ( Convolutional Neural Network) ． In test phase，the coefficients of random noise can be predicted from the coefficients of desert seismic data by the mapping relationship，and thereafter the effective signals coefficients is obtained indirectly． Finally the effective signals can be reconstructed by inverse Shearlet transform． By comparing with the traditional Shearlet hard threshold denoising algorithm，the proposed algorithm has improved the SNＲ of the desert seismic signals from － 4． 48 dB to 14． 15 dB and has better denoising performance．

The gap between the mutual information of the Gray-coded amplitude phase shift keying ( APSK) and the channel capacity can not be narrowed sufficiently，especially with non-iterative independent demapping at the receiver，in order to solve the problem，a non-equiprobable APSK constellation is proposed． The mutual information under the restriction is improved by reducing the number of constellation points corresponding to the inner loop and transforming it into a low-order Gray-PSK． Simulation shows that the 64NE-APSK designed for 2 /3 code rate has a performance advantage of 0． 6 dB compared to the 64Gray-QAM constellation，having an advantage of about 0． 1 dB over the 64Gray-APSK． Therefore，in the case of independent de-mapping at the receiver，the design helps approaching the channel capacity without increasing the complexity of the system implementation．

There are strong couplings between the attitude motion of a hypersonic vehicle，which can be divided into motion coupling，aerodynamic coupling，and inertial coupling． The aerodynamic coupling of aircraft is analyzed，the attitude controller is designed to weaken aerodynamic coupling． Firstly，the relationship between the aerodynamic moments of the hypersonic vehicle's three channels is analyzed． Secondly，by using the coupling analysis method of dynamic equations，the coupling relationship between the channels is obtained and the coupling matrices are solved． Then，combining the coupling matrices to judge whether coupling can be ignored，an attitude coordination controller is designed． The simulation results show that the designed controller can effectively weaken the aerodynamic coupling between the three channels of the aircraft，and improve the dynamic response ability of the system on the premise of ensuring the stability of the attitude．

In uncertain networked control systems with distributed delay，in order to solve the problem of the delay and packet loss in the network channel and the parameter perturbation of the controller，etc． ，a design method of non-fragile L2-L∞ controller with distributed delay and network-induced delay is proposed． By selecting a delay-dependent Lyapunov function and constructing the free weighting matrix， the L2-L∞ performance criterion to guarantee the asymptotic stability of closed-loop control systems is derived． The design problem of robust L2-L∞ controller is converted into a convex optimization problem of LMI ( Linear Matrix Inequality) by using the LMI technique． Finally，the comparisons of numerical simulation results verify that the controller can make the open-loop divergent system asymptotically stable under the given performance index，and have better convergence effect than the conventional controller without parameter perturbation．

A coordinated control system consisting of three PMSLM ( Permanent Magnet Synchronous Linear Motors) is taken as the research object，and the effects of different topological structures on the synergy performance between the motors are studied． An adjacent offset coupling cooperative algorithm is proposed to improve the synergy，the overall control effect of the system． The experimental results show that after the application of adjacent offset coupling algorithm，the accuracy of collaboration between PMLSM is improved． The collaborative error between adjacent motors is reduced from 0． 3 mm to 0． 25 mm，and the accuracy of collaboration is improved by 17%．

To improve the compatibility of operating system kernel function call relations generation，a generation method based on hardware virtualization is proposed． This method replaces the kernel instructions in specific position with special instructions． These special instructions trigger virtualization interrupt trapping when the function is calling or called． So the virtual machine monitor can dynamically generate the operating system's kernel calling relations in the interrupt trapping． Experimental results show that this method can generate either open-source or closed-source，32 bit or 64 bit operating system's kernel function calling relations without relying on kernel source code or compilers． The accuracy of the generation is 100% and recall is 85%． This method can be used in the operating system kernel security analysis and whitelist generation．

In order to solve the problem of inefficiency of traditional bird-driving methods at the Airport，this paper introduces the technology of multi-UAVs cooperation into this field． We study a bird-driving technology by using multi-UAVs cooperative． Based on the self-created“bird-ambushed”strategy，a mission planning model is built and a task assignment method is designed based on genetic algorithm． In order to evaluate the parameters of bird strike threat，the UAV interception efficiency which are the preconditions of the mission planning． The reasoning tree analysis and fuzzy logic analysis were introduced to optimize the parameters，respectively． In order to ensure the actual effect of mission planning，multiple constraints are designed，and the mission time of UAV is estimated based on proportional guidance law． The numerical simulation results show that，under the condition of multiple constraints and multiple bird targets simultaneously，the developed method can effectively accomplish the work of multi-UAVs cooperative bird-driving．

In order to promote the development of path planning technology，the planning ideas and forms of path planning are analyzed． The path planning algorithms are divided into the traditional classical algorithms and modern intelligent algorithms in two categories，and some commonly used algorithms are analyzed and summarized．And the current research hotspots and future development trends are pointed out from the three aspects of improving the application of modern intelligent algorithms in path planning，amalgamation of multiple algorithms and the research of four-dimensional path planning algorithms for multiple UAVs．

In order to compare similarity of pure vowels in Korean and Mongolian using computational phonetics method，this paper proposes an approach of comparative analysis based on formant parameters． Firstly，F1 and F2 formant frequency is extracted from Korean and Mongolian pure vowels，the mean value，standard deviation and divergence factor of F1 and F2 parameters are computed respectively． Secondly，analyze the distributing differences of formant frequency for each of the two languages are analyzed． Finally，similarity of Korean and Mongolian pure vowels are compared by t-test method with formant frequency． Comparative analysis results show that the divergence between different acoustic features of pure vowels can be clearly distinguished by F1 and F2 ． And the total 5 pairs pure vowels are similar in acoustic features，and the 5 pairs pure vowels are in correspondence with analysis of tongue-position and lip-rounding in pronunciation process． The proposed approach can provide a basis for further study on similarity between Korean and Mongolian by using computational phonetics．

In the global feature extraction and recognition of finger knuckle print based on PCA ( Principal Component Analysis) ，in order to improve the accuracy，we prove that the knuckle patterns of four fingers have different classification weights，i． e． contribution rate to identification results according to researches on traditional methods． This way improved the equal weight distribution method shown in existing research consequences． The experiment results show that when the four-finger weight ratio is 2 ∶ 2 ∶ 3 ∶ 2，the maximum recognition rate，which is 94． 4%，can be achieved when 21 dimensional characteristics are used． In contrast，94． 4% can be achieved when 32 dimensional characteristics are used in the equal weight distribution．Therefore，this method can reduce feature dimensions and improve the recognition speed．

In order to solve the energy loss problem of the new energy connected to the large power grid，economic dispath is an effective method to realize fast economic dispatch of generating units with conventional energy． Considering the existence of power systems with multiple energy sources，an economic dispatch model of power systems with transmission loss is established． Based on the extended Lagrange-Hopfield neural network optimization algorithm，the multiplier neuron transmission rate parameter is introduced，and the improved extended Lagrange-Hopfield neural network algorithm is proposed． The convergence of the algorithm is proved theoretically，and the algorithm has fast convergence speed． Taking a project as an example， different transmission rate parameters are selected to verify that the property of the improved augmented Lagrange-Hopfield neural network algorithm is affected by transmission rate parameters． The method is feasible in engineering application and can converge rapidly．

Improving screening and addressing terrorist information in social media becomes a critical topic of research． Our study uses data crawlers and screening analysis to classify different terrorist information． Through scrappy framework and genism tool pack，we obtained the relevant data from four datasets． After constructing a LDA ( Latent Dirichlet Allocation) theme model，we then classified several themes in four different area that played major roles in propagating terrorism，and investigated the interconnected relationship between these themes． The experiment indicates that the model can effectively distinguish different themes of terrorism information and the role it played in propagation．

Alzheimer's disease is a common disease and is harmful to the elderly． At present，there is no effective cure for Alzheimer's disease． It can only reduce the harm of Alzheimer's disease to the patient，family and society through nursing． Alzheimer's disease is a chronic disease and the nursing cycle is long，so it often causes great physical and mental stress to the nursing staff． In order to reduce the patient's pain and relieve the burden of the caregivers，an intelligent wearable device based on signal acquisition and wireless communication is developed． On the one hand，it can monitor the patient's behavior in real time． The patient's foot pressure changes are collected by PVDF ( Polyvinylidene Fluoride) sensors，then the behavior is notified and the caregiver is notified to ensure patient safety． On the other hand，we can locate the patient's position information by GPS( Global Positioning System) and notify the caretakers in real time by SMS ( Short Message Service) ，and the device has the function of emergency notification． This smart wearable device adapting to Alzheimer＇s patients can help the caregivers take care of the patients and ensure the safety of the elderly．

In order to realize the quadrotor's automation and informatization in transportation，we designed and realized a quadrotor which could track and grasp target automatically and then carry it to the specified place．By applying the control theory of quadrotor，the quadrotor could be able to judge its position in real time using computer vision technology and acquire its amplitude with a sonar sensor，thus adjust its flying amplitude and attitude angel with the PID ( Proportion-Integral-Derivative) controller． Thus，the system finished automatically tracking and target identification using camera and drove a manipulator to accomplish target grasping and releasing． The result shows that the quadrotor could keep its flying height at 1 m，and fly at the speed of 20 cm /s． The machine could recognize and grasp a cylindrical target of which the radius is 4 cm and the length is 8 cm． The whole process can be finished in 50 s．

Multilayer perceptrons of convolutional neural networks use the gradient descent algorithm for training，there are often problems of slow convergence and small localization． In order to solve the problem，we propose a method to calculate the optimal initialization weight by using information interaction to improve the proposed network structure，which can effectively reduce the training time and avoid the local minimum problem． Firstly we use mathematical theory to derive the formula for the optimal initialization weight of the ＲeLU function． We target 2-channel network structure using this method，directly substituting data to find the optimal initial weight．Through multiple training and testing of the three data sets，better results are obtained． Average matching accuracy for grayscale images is increased by about 1． 5%． For the FPＲ95，the average value also dropped from 5． 23 to 4． 65． Initialization weight setting prevents neurons from entering the hard saturated region． And it has the advantages of higher precision，stable effect and fast convergence．