The traditional speech-based Parkinson's Disease (PD) prediction method is to predict the motor Unified Parkinson's Disease Rating Scale (motor-UPDRS) and the total Unified Parkinson's Disease Rating Scale (total-UPDRS) respectively. In order to solve the problem that the traditional method could not use the shared information between tasks and the poor prediction performance in the process of single task prediction, a multi-task regression method based on model filtering was proposed to predict the motor-UPDRS and total-UPDRS of Parkinson's disease patients. Firstly, considering the different effects of the subtask speech features on the predicted motor-UPDRS and total-UPDRS, an L1 regularization term was added for feature selection. Secondly, according to different Parkinson's patient objects distributed in different domains, a filtering mechanism was added to improve the prediction accuracy. In the simulation experiments of remote Parkinson data set, the Mean Absolute Error (MAE) of motor-UPDRS is 67.2% higher than that of the Least Squares (LS) method. Compared with the Classification And Regression Tree (CART) in the single task, the motor value increased by 64% and the total value increased by 78.4%. The results of experiment show that multi-task regression based on model filtering is superior to the single task regression algorithm for UPDRS prediction.

Concerning the problem of low convergence accuracy and being easily to fall into local optimum of the Particle Swarm Optimization (PSO), an improved PSO algorithm combined Centroid and Cauchy Mutation, namely CCMPSO, was proposed. Firstly, at the initialization stage, chaos initialization was adopted to improve the ability of initial particle uniform distribution.Secondly, the concept of centroid was introduced to improve the convergence rate and optimization capability. By calculating the global centroid of all the particles in the population and the individaual centroid formed by all of the individuals' extreme values, sufficient information sharing could be realized in the interior of the particle swarm. To avoid falling into local optimal solution, Cauchy mutation operation was used to perturb the current optimal particle, in addition, the step length of disturbance was adaptively adjusted according to the operation rule of Cauchy mutation; the inertia weights were also dynamically adjusted according to population diversity. Finally, seven classical test functions were used to verify the algorithm. Experimental results indicate that the new algorithm has good performance in convergence precision of the function execution results, including the mean, the variance and the minimum value.

Traditional power allocation schemes have ignored channel estimation errors and circuit energy consumption. To solve this problem, an improved green joint optimization scheme was proposed in this paper. On the premise of guaranteeing user's QoS (Quality of Service), energy efficiency and time efficiency for relay selection and each relay's power allocation were jointly optimized in the improved scheme, with taking channel estimation errors and relay circuit energy consumption into consideration. In the end, the closed solutions of transmit power of the source node and relay nodes were obtained. The simulation results show that the proposed scheme performs 30% better than traditional optimization scheme in energy efficiency with high SNR (Signal-to-Noise Ratio) and has a close performance with low SNR.

To correctly obtain the high-level language control structures of embedded executables and assembly code, and resolve the problem that the existing recovery methods for high-level language control structures cannot handle the unstructured region, the classical control analysis method, structural analysis algorithm, was introduced to study the recovery method for high-level control structures of embedded assembly code. The structural analysis algorithm was improved according to the characteristics of embedded executables, and the high-level language code was generated by using the program control tree, which can be obtained from the results of structural analysis algorithm. Compared with the open source decompiler named DCC, the results show that the improved algorithm is feasible and efficient.

Two-stage cooperative spectrum sensing that used both soft combining and hard combining need to send much unnecessary information to fusion center in the second stage. Concerning this problem, this paper proposed an improved algorithm by estimating Signal-to-Noise Ratio (SNR). The improved algorithm selected traditional two-stage scheme to improve the performance of spectrum when SNR was low and selected hard combining scheme to decrease the number of sending data when SNR is high. The simulation results show that the proposed algorithm decreases much unnecessary soft information being sent to fusion center at the cost of a little loss in performance of detection as compared to an existing two-stage cooperative spectrum sensing algorithm.

Concerning the application requirements of slope stability analysis in large areas and the defects of current calculation method, a three-dimensional analysis method based on Digital Elevation Model (DEM) data was proposed. In this method, a sphere was used, instead of ellipsoid, to search slippery surface, and the slope safety factor under the three-dimensional conditions was calculated through integral operator with the results of two-dimensional analysis. Finally, the position and shape of possible landslide were determined according to the safety factor. The practical application results confirm that this method simplifies the search algorithm, ensures the accuracy of slope stability analysis, and improves the efficiency of analysis and calculation.

An automatic annotation method for professional literature was proposed. Through comparing with other storage formats and literary styles, two features of professional literature were summarized, and then three assumptions were proposed. To improve annotation efficiency, based on topology structure, the domain ontology was partitioned into segments which were self-consistent, then the most related segments were located with the keywords extracted from document, finally the document with located segments was annotated and the annotation scope was expanded according to the correspondence between grammatical structure and semantic structure. The experimental results show that the proposed method can improve annotation efficiency, annotation quantity and annotation accuracy.