The Research And Development Of Mobile Dictation Software Based On Image Recognition

With the acceleration of global informationization, it is a must to improve the efficiency of information processing. As Android operating system and the research of Optical Character Recognition are becoming more and more popular, it is becoming a hot topic to process and store the document information efficiently by using smart mobile devices.The paper analyzed the characteristics of mobile images, and studied algorithms used in OCR, like character preprocessing, feature extraction and classification recognition. In character preprocessing, source images are processed from being compressed, gray processing, binarization processing, direction correction, character segmentation and normalization. The source images are compressed proportionally by the bilinear interpolation algorithm. OpenCV library is used to do image gray processing. Comparing with the traditional method of image binarization, the paper studies and designs different binarization processing for different images and enhances the traditional Bernsen algorithm. This paper calculate the source image and Gaussian filtered image by Bernsen, and recalculate two different thresholds to get the new threshold. Run Length Smoothing algorithm and image thinning method are used to adjust slanting angle of the slanted pictures Character pictures are segmentated by horizontal projection and vertical projection into each character image, and then normalized to 24*36 px. After studying both the structural and statistical feature extraction methods, the paper comes along with the feature extraction solution:54 grid features and 20 penetration features and the combined features are written to file. With the combined feature, a three-layer BP neural network algorithm is used to identify the characters. At last, the paper does a simulation experiment of neural network, reads and classifies the feature file, and finally proves the feasibility of this algorithm with the experimental results.According to the research and the requirements of the English words dictation, the primary and secondary school English words dictation system is designed. The system is based on Android operating system. The dictation software mainly consists of three parts, which are image recognition, words dictation and system functions respectively. Dictation methods include phone dictation and paper dictation. The software accomplishes the change from image to text by OCR,and text to voice by Text To Speech. It can directly and easily recognize and read the English words from the images, which can help the teachment and words dictation.