The key of lip-reading recognition is to analyze rules of lip shapes, and build up the corresponding relationships between lip shapes and Chinese speeches, but the difficulty of the research lies in the "one-to-many" relationship between lip shape and speeches. For this problem, the associational lip-reading recognition based on Bayesian theory was put forward by this paper. The method applied Bayesian theory to recognize Chinese speeches, simultaneously, dynamically and self-adaptively adjusted the probabilities of all speeches corresponded by the same lip shape based on an award and penalty strategy, consequently tried to reflect speaker's real intention to the greatest extent. The results verified that this method could recognize effectively Chinese characters having the same lip shape, and enhanced the lip-reading system's practicability and validity.
In this paper,the convergence analysis for a new preconditioned Gauss-Seidel iterative method was discussed.If the coefficient matrix is a nonsingular irreducible M-matrix,the convergence rate of this iterative method depends on the spectral radius of the original Gauss-Seidel method.Likewise,the spectral radius of the preconditioned Gauss-Seidel iterative method also depends on one of the preconditioned Gauss-Seidel methods.Finally,some numerical examples are given to explain our theoretical results.
In this paper,we discuss the feasibility of application of machine learning algorithm in teaching quality evaluation,and use weighted margin support vector machine(WMSVM) model in learning many metrics of teaching quality evaluation in order to build up a stable model to evaluate teaching quality.Teaching quality metrics are input to the model and confidence value of attitude information of students.In our model,this confidence is expressed as weighted margin between the sample points and the classification hyper plane.The WMSVR model outputs teaching quality metric as scores.Comparing these scores with experts' evaluation,we find that WMSVR model is more accurate and of better generalization capability in the procedure of teaching quality evaluation.
For the resource limit of multimedia data processing in embedded device,an open source video server software MJPG-streamer to implement the processing and storage of multimedia data is used.The source code is modified and a new plug-in in this software to enhance its function is added,based on the features and architecture of this software.Transplant it to the embedded systems after cross-compile,to implement a video surveillance system based on Arm9,which also can control a front-end equipment.The experimental result shows that this method can enhance the function of video data processing,and the video can be played fluently in client browser.
In this paper,on the base of the property for the selection of set-valued mapping,we introduce a relation of almost lower semicontinuous and continuous selection for metric projection in Banach space.
Based on the two-TFT pixel driving circuit for AM-OLED, 4-TFT driving circuits which include current-programmed and voltage-programmed were analyzed. Although they could compensate the threshold voltage variations, each of them has its advantages and disadvantages: the former could control the scale of the brightness well, but its disadvantage was slow response; the latter has fast response, but it could not control the scale of the brightness well.
We introduce symmetric generalized quasi-vector equilibrium problems. Moreover, through nonlinear scalarization function, together with a fixed-point theorem, we prove existence theorem of solutions for symmetric generalized quasi-vector equilibrium problems. The results in this paper improve some recent results.
To increase the efficiency of elevator group system, the paper introduced a new designed simulation system based on non-stationary Poisson process and Model/View/Controller (MVC) pattern. At the same time, a new dispatch algorithm was proposed based on machine vision to inspect the hall traffic. The proposed algorithm can serve one hall-call by allocating more than one elevator. Finally simulation results show that the new method outperforms the minimum long-waiting method in terms of average waiting time and left-waiting people.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)