水下机器人三维路径制导控制器设计

本文从水下机器人的具体结构出发，根据视线制导控制规律推导了水下机器人的一个三维变结构路径跟踪控制器。仿真结果令人满意。     

Clinical Detection and Feature Analysis on Neuro Signals

Research on neuro signals is challenging and significative in modern natural science. By clinical experiment, signals from three main nerves (median nerve, radial nerve and ulnar nerve) are successfully detected and recorded without any infection. Further analysis on their features under different movements, their mechanics and correlations in dominating actions are also performed. The original discovery and first-hand materials make it possible for developing practical neuro-prosthesis.     

AN EFFICIENT FAST ENCODING ALGORITHM FOR VECTOR QUANTIZATION

IntroductionAs an effective technique for data compression,VQ[1,2 ] has been successfully used forvarious ap-plications involving VQ- based encoding and VQ-based recognition in both vocal and image pro-cessing.The k- dimensional,N- level vectorquantizer is defined as a mapping from a k- di-mensional Euclidean space Rkinto a certain finitesubset C={Ci,i=1 ,2 ,… ,N },which is called acodebook and whose elements Ci are called code-words.A distortion measure D( X,Ci) is a non-negative dissi…     

Orthogonal discriminant improved local tangent space alignment based feature fusion for face recognition

Improved local tangent space alignment (ILTSA) is a recent nonlinear dimensionality reduction method which can efficiently recover the geometrical structure of sparse or non-uniformly distributed data manifold. In this paper, based on combination of modified maximum margin criterion and ILTSA, a novel feature extraction method named orthogonal discriminant improved local tangent space alignment (ODILTSA) is proposed. ODILTSA can preserve local geometry structure and maximize the margin between different classes simultaneously. Based on ODILTSA, a novel face recognition method which combines augmented complex wavelet features and original image features is developed. Experimental results on Yale, AR and PIE face databases demonstrate the effectiveness of ODILTSA and the feature fusion method.     

Overlay Share Mesh for Interactive Group Communication with High Dynamic

An overlay share mesh infrastructure is presented for high dynamic group communication systems, such as distributed interactive simulation (DIS) and distributed virtual environments (DVE). Overlay share mesh infrastructure can own better adapting ability for high dynamic group than tradition multi-tree multicast infrastructure by sharing links among different groups. The mechanism of overlay share mesh based on area of interest (AOI) was discussed in detail in this paper. A large number of simulation experiments were done and the performance of mesh infrastructure was studied. Experiments results proved that overlay mesh infrastructure owns better adaptability than traditional multi-tree infrastructure for high dynamic group communication systems.     

An Adaptive Scheme for Neighbor Discovery in Mobile Ad Hoc Networks

The neighbor knowledge in mobile ad hoc networks is important information. However, the accuracy of neighbor knowledge is paid in terms of energy consumption. In traditional schemes for neighbor discovery, a mobile node uses fixed period to send HELLO messages to notify its existence. An adaptive scheme was proposed. The objective is that when mobile nodes are distributed sparsely or move slowly, fewer HELLO messages are needed to achieve reasonable accuracy, while in a mutable network where nodes are dense or move quickly, they can adaptively send more HELLO messages to ensure the accuracy. Simulation results show that the adaptive scheme achieves the objective and performs effectively.     

FAST IMAGE ENCODING ALGORITHM BASED ON MEAN-MATCH CORRELATION VECTOR QUANTIZATION

Introduction　　 Vector quantization (VQ) is an importanttechnique for image compression,and has beenproven to be simple and efficient[1,2 ] .Signal com-pression is achieved in VQ by using a codebookwith relatively few codevectors compared to thenumber of possible signal vectors. VQ can beviewed as a mapping from k- dimensional Euclideanspace Rk into a finite subset C of Rk.We call thisfinite set C the codebook and,moreover,C={ci|i= 1 ,2 ,… ,N},where ci is a codeword and N is thecodeboo…     

Feature Selection for SVM Classifiers Based on Discretization

The rough sets and Boolean reasoning based discretization approach （RSBRA） is not suitable for feature selection for machine learning algorithms such as neural network or SVM because the information loss due to discretization is large. A modified RSBRA for feature selection was proposed and evaluated with SVM classifiers. In the presented algorithm, the level of consistency, coined from the rough sets theory, is introduced to substitute the stop criterion of circulation of the RSBRA, which maintains the fidelity of the training set after discretization. The experimental results show the modified algorithm has better predictive accuracy and less training time than the original RSBRA.     

Segmented dynamic optimization model for asset-liability management of commercial banks and its applications

Asset-liability management is the core business of commercial banks. Effective method of asset-liability management is a continuously exploring topic in the academic and practical fields. According to the operational characteristics of commercial banks, this paper addresses a segmented dynamic optimization model under the perspective of the regulatory environment for China commercial banks. The model can perform segmented sliding optimization and correct control variables to make optimal decision with the changes in situations for a certain future time.     

RSOFCPN: CONTROL SYSTEM STRUCTURE AND ALGORITHM DESIGN

IntroductionThe learning approaches of neural network canusually be distinguished into three classes:su-pervised learning,unsupervised learning and re-inforcement learning.The supervised learningscheme has the highest efficiency but requiresthe learning systems to provide with desired out-puts.Obviously,it is very difficultin many cas-es.Unsupervised learning methods constructin-ternal models at capture regularities in their in-puts without receiving any additional informa-tion.Therefore,its …