Research Of Medical Image Information Collection System Based On Android Terminal

With the decelopment of the society, people pay more and more attention to their heathy condition, and it is necessary to monitor the medical physiological data of the user in the real time. Conventional wisdom medical approach collect data through wireless mode and send them to the remote server. In this paper, a new intelligent medical scheme is proposed. First, human physicological data can be collected, then display the data on the seven segment code Liquid Crystal Display screen terminal. Last, store the result in the database and upload to the server after identifing the recognition technology with the image which captured by Android smartphone.In this paper, the image recognition achieved in two ways: one is the Back Propagation neural network method which needs for image pre-processing and feature extraction. Then the characteristics are identified by the classifier; The other is taking the whole graph as the input feature of image recognition with the convolutional neural network.The main research work and innovation points of this paper are as follows:(1) A medical physiological system is designed which include data acquisition module and display module. The system takes MSP430 microcomputer as the control chip and display the physiological data which are collected by sensor chip on the LCD screen;(2) After analysing the principle of BP neural network, some improved method is put forword. A large number of LCD seven segment code of digital image samples are collected and then the recognition classifier can be got after being trained with visual stdio 2010 platform. The test shows that the recognition rate can reach 98%.(3) Android client application is Designed. Its features include image acquisition, user data storage, image recognition and data storage, etc. The image preprocessing and feature extraction can be realized with the android native interface function. It recognizes the image with the trained neural network classifier. The physiological data can be saved in the database for checking.(4) Combined with the deeplearning technology the handwritten digital library MNIST and the seven segment code of the LCD screen can be identified and the result shows its recognition rate is higher than that of BP neural network. A weight reduction method is proposed for facilitating the deep network. The experimental results shows that the parameter weights of network can be reduced by 82% and still keep a high recognition rate.