Computational Analysis of Handwriting, Character Recognition, and Design of Digital Fonts

在过去的几十年里，我们已经探索和开发了许多识别不同类型模式的计算技术，包括但不限于不同语言的笔迹和多个年龄组的人脸。这项发现资助申请涉及发现新的基于知识的特征的新技术，降低错误率和优化不同识别算法组合的方法，以及识别用英语，法语，阿拉伯语和其他语言书写的字符和单词。它的目的是提高我们的识别算法，使用多阶段的多分类器，结合我们的速示器和眼动跟踪实验中发现的新功能，在过去的几年中。它将专注于我在以下新领域的研究扩展，并与笔迹学和心理学专家以及字体设计师进行磋商：- 提取更详细的独特笔迹特征，例如字符的形状，字符的形成方式以及个人书写的特点，以验证笔迹与个性特征之间的相关性，这在笔迹学中是一个具有挑战性和争议性的话题，并检测书写者的健康状况，例如患有帕金森病和阿尔茨海默病的书写者。我们相信，我们在手写识别方面的丰富经验将使我们能够为这一主题带来新的见解，并为笔迹学家节省分析个人笔迹的时间提供帮助。- 应用我们的字符识别和感知技术，评估汽车牌照中使用的字体的易读性，并提出新的字体设计，以便于通过图像处理和模式识别技术识别车牌，以帮助执法机构查找被盗汽车，交通违法行为，在不同天气条件下识别车辆以及无人驾驶车辆的指挥。- 将我们的特征提取技术，如硬币上铸造的字符或纸币上印刷的字符的关键点的提取，扩展到检测假硬币和纸币，这已经成为一个迫在眉睫的重要课题，影响到整个世界各国的经济。 - 对非常大的数据集进行密集的实验，以发现帮助我们社会的社会和经济福祉的新方法。

Over the past decades, we have explored and developed numerous computational techniques of recognizing different types of patterns, including but not limited to handwriting in different languages, and human faces of multiple age groups. This Discovery grant application is related to novel techniques of discovering new knowledge-based features, methods of reducing the error rate and optimizing the combination of different recognition algorithms, and the recognition of characters and words written in English, French, Arabic, and other languages. It is aimed at improving our recognition algorithms using multi-stage multiple classifiers by incorporating the new features discovered in our tachistoscope and eye-tracking experiments during the past years. It will be focused on an extension of my research in the following new areas in consultation with experts in graphology and psychology, and font designers: - Extract more detailed distinctive handwriting features, such as the shapes of characters, the way characters are formed, and peculiarities of the writing of the individuals, to validate the correlation between handwriting and personality characteristics, which is a challenging and controversial topic in graphology, and to detect the health conditions of the writers, e.g. those with Parkinson and Alzheimer diseases. We believe our rich experience in handwriting recognition will enable us to bring new insights to this subject and offer help to graphologists in saving their time to analyze the handwriting of individuals. - Apply our character recognition and perceptive techniques to assess the legibility of the fonts used in the license plates of automobiles and propose new designs of fonts to facilitate the recognition of the license plates by image processing and pattern recognition techniques to help law-enforcement organizations in finding stolen cars, traffic offences, identification vehicles under different weather conditions and the command of driverless vehicles. - Extend our feature extraction techniques, such as the extraction of keypoints of the characters minted on the coins or printed on banknotes, to detect fake coins and paper money, which has become an imminent subject of importance affecting the economy of all countries around the whole world. - To conduct intensive experiments on very large sets of data to discover new ways of helping the social and economic well-being of our society.