Android 中通过模式匹配的离线手写识别应用程序
我正在考虑使用神经网络开发一个用于离线手写识别的应用程序。我想开发一个应用程序，它可以帮助我首先通过接收用户的模式并将其与某些字母表相匹配来训练系统。然后使用这个经过训练的系统进行手写识别。我想知道如何保存图像以及如何将它们映射到角色？将它们存储为单独的图像并维护数据库以将它们与相应的字母表相匹配将使应用程序变得相当繁重。对于存储模式并随后匹配它们还有其他想法吗？对 android 中的存储和后端活动不太熟悉:(
I'm thinking about developing an application for offline Handwriting Recognition using neural network. I want to develop an application that will help me first train the system by taking in patterns from the user, and matching them to certain alphabets. And then use this trained system for Handwriting Recognition. I want to know how to save the images and how to map them to characters? Storing them as individual images and maintaining a database to match them to their corresponding alphabets will make the application quite heavy. Any other ideas for storing patterns and later matching them? Not much acquainted with storage and backend activities in android :(
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（2）
您可能想查看“Dr. Dobb's Journal”举办的手印识别竞赛的结构和结果：
http://drdobbs.com/security/184408923?pgno=2
You might want to examine the structure and results of the Handprinting Recognition contest, held by "Dr. Dobb's Journal":
http://drdobbs.com/security/184408923?pgno=2
有不同的方法可以实现这一点，但通常的想法是以某种矢量格式存储用户输入。一种简单的方法是仅存储用户在屏幕上触摸的点的坐标，可能会将它们标准化，以便它们不依赖于物理屏幕尺寸（例如，重新映射到虚拟 400x400 屏幕等）
There are different ways to implement this, but the usual idea is to store the user input in some sort of vector format. An easy way is to store just the coordinates of the points the user touches on the screen, possibly normalizing them so they are not dependent on physical screen size (e.g., remap to a virtual 400x400 screen, etc.)