文章摘要

文章信息/Info

Author(s):: ZHANG Jing-lin; WANG Xu-zhi; WAN Wang-gen; WU Yong-liang; School of Communication and Information Engineering,Shanghai University,Shanghai200072，China

Keywords:: handwriting recognition; pre-processing of character recognition; Bresenham algorithm; entropy power

摘要:: 为了使语言障碍者克服障碍，通过电话等现代通信手段与其他人进行语言沟通，设计了一套实用且低成本的手写识别语音播报系统。语言障碍者在触摸屏上书写汉字字符，书写轨迹坐标经过处理转化为二值图像，再经过坐标数据的平均化、平滑处理、非线性规范、改进的Bresenham算法等预处理操作，最后调用基于熵功率的字符识别算法。字符识别完成后，将识别得到的字符编码分帧发送至TTS语音器件输出模拟音频。实验证明该系统识别准确率较高，平均达到92%，接近实际应用标准，因此该系统具有一定实用性和市场价值。

Abstract:: In order to help the mute people to overcome the obstacle and exchange ideas with others by telephone and other modern communication ways,a research of low-cost pronunciation-broadcast device with the function of handwritten character recognition is presented in this paper.Firstly,the DSP transfered the writing track into the binary image after the mute people wrote words on the touch screen,then pre-processing of character recognition were necessary,such as equilibration of coordinate,smoothing,non-linear specification of coordinates,Bresenham algorithm and so on.The DSP applied the function of character recognition to complete the recognition of characters.The recognized characters were coded and sent to the TTS chip.Through the analysis of the text and speech synthesis,the chip outputed analog audio with correct pronunciation.The experimental results testify that this system shows a high recognition rate and the average reaches92%.So the system has a certain utility and market value.