Design And Development Of Embedded Blind Reader

In our daily life, most of the information which we are access to is passed in the form of image and it will lead to a serious result that the blind and visually impaired people could not get most of the important information without other people’s help. All these will reduce the chance of the blind to perceive the world. The blind reader is such a device which achieves the conversion from image to voice signal combining the TTS and OCR technology. This device can make people with visual impairments get information from paper and electronic edition text more freely.This paper chose QT to complete the GUI programming of the blind reader software. The software can be divided into the following three parts.(1) The video capturing and image pre-processing module. This module mainly does the capturing of image which is waited recognition and the pre-processing which includes graying, binarization, denoising, layout analysis, slant correction and character segmentation. Through comparative analysis, this paper selected the suitable algorithm and has obtained good results. This module is the foundation of the whole system and its performance is directly related to the accuracy of recognition.(2) The text recognition module. This module is mainly composed of character preprocessing, feature extraction and text recognition. It is the core of the whole system. This paper chose appropriate thinning and normalization algorithm to complete character preprocessing and introduced some widely used features extraction process of Chinese characters. Considering that Chinese characters recognition belongs to large category pattern recognition, this paper designed multi-layer classifier based on multi-features. Coarse classification has narrowed down the scope of character set as much as possible and reduced calculation in the process of fine classification effectively. Various features were combined to distinguish the right character in the fine classification process which leaded to greatly improvement of the recognition accuracy.(3) The speech broadcast module. Because of the different service environment, three methods of speech broadcast were implemented in this paper including the way based on Microsoft Speech SDK, IFLYTEK MSP and homemade audio library. All these have got good success. The way of homemade audio library is suitable for various platforms, but the speech is not fluent and it also takes a larger memory space for the audio library.In addition, this paper added the speech recognition module for increasing the real-time interactivity of the Windows software. So our software can receive user’s speech as input and deliver information to user with speech. The result shows that this system realized the basic function of blind reader intelligently and friendly. The accuracy of Chinese character recognition reaches 99.57%. So the software has reached the design requirements.Lastly, this paper implemented the development of blind reader software based on the QT embedded Linux system. The embedded system chose SEED-DVS6446 as hardware platform and QT/Embedded as embedded GUI development tool which make it convenient for debugging. This paper has introduced the process of building the embedded Linux development environment in detail, including the construction of cross compile environment, configuration and transplantation of kernel, the transplantation of QT/Embedded and so on. Finally, we can get the executable file of the embedded blind reader application by means of cross-compile and then run the program on the target hardware platform through the NFS network file system. After commissioning and repeated efforts, this system has achieved the function of blind reading and ran well.