The Research And Implementation Of Real-time Translation System Based On Mobile Cloud

As the development of computer science and electronic technology, the mobile phone is not only a tool of communication, but also an open platform. It makes people’s life much more convenient by abundant amount of applications. Android system is a smart operating system released by Google in 2007, and its open and powerful features make it easier to develop mobile applications. Along with the growth of material and cultural requirement, people’s need of mobile applications increases continuously. The development of applications has a great prospect.The thesis focuses on the daily need of people and develop a real-time translate system based on mobile cloud. The application can add the translation results in the real-time scenes like augmented reality. Then we improve its performance by using mobile cloud due to the limited computing resource and uncertain runtime environment. The traditional cloud computation fixes the tasks that need heavy computation in the cloud, it cannot get the best performance in some extreme environment. At the same time of implementation of real-time translation, this thesis proposes a new offload plan by cooperation of mobile and cloud. It can get better performance due to the actual runtime environment.Concretely, the main parts of this thesis conclude:1. Make a decision of the existing methods of image progressing, then optimize the algorithm of filling holes and labeling connected domains. Introduces a research on locating the text areas from image with complex background based on morphology and connect domain analysis.2. Based on the result of text location, tailor the text area of the image and then recognize the image to texts. Translate the English texts and combine the results of translation with the original image.3. Introduce a new dynamic plan of cloud offloading. Differ from the traditional cloud computation, this plan divides the program into several tasks, the tasks can run in either the local phone or cloud by the cooperation of both sides. By the real-time analysis of running environment information, historical running data and future resource estimation, the plan dynamically decides the running place of each task. It gets the best running performance by choosing the running plan with lowest consumption of time or energy.At last, this thesis shows the achievement of real-time translation and the most beneficial way chosen by the dynamic offload plan. Result shows the system has achieved the purpose.