HHreco Software Download Page

HHreco is a software library providing multi-stroke symbol recognition and segmentation utilities written in Java. In addition to interfaces and core classes for developing gesture recognizers, the library contains a solid implementation of an adaptive multi-stroke recognition system. The recognition algorithm is invariant to stroke-order, -number, and -direction, as well as invariant to rotation, scaling, and translation of symbols. It is designed to be an off-the-shelf recognizer and can also be further customized to specific applications.

The detailed description and the performance evaluation of this recognition system have been documented in the following technical report:

Heloise Hse and A. Richard Newton, Sketched Symbol Recognition using Zernike Moments, Technical Memorandum UCB/ERL M03/49, Electronics Research Lab, Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, CA 94720, Dec 2003. [pdf]

Software tutorial:

Heloise Hse and A. Richard Newton, Graphic Symbol Recognition Toolkit (HHreco) Tutorial, Technical Memorandum UCB/ERL M03/50, Electronics Research Lab, Department of Electrical Engineering and Computer Sciences, University of California, Berkeley, CA 94720, Dec 2003. [pdf]

Downloads:

Jan 28, 2004 -- HHreco, hhreco.tar.gz (1.4MB)
This is a precompiled package including all .java files, .class files, and the API documentation. It comes with the recognition utility and also the symbol segmentation utility.

May 18, 2004 -- Symbol Database, data.tar (16MB)
This tar file contains the test corpus collected from 19 different people and each file can be parsed using MSTrainingParser included in HHreco. Each user's data file contains at least 30 examples per symbol class (see data statistics). Our experiments are performed using the first 30 examples of every symbol class of every user's file. To date, there are a total of 7791 multi-stroke examples overall and at least 575 examples per symbol. The symbol set consists of the following shapes:

Please reference our work when you are using the software, database, and documentation distributed on this website.

Publications:

Helose Hse and A. Richard Newton. "Robust Sketched Symbol Recognition using Zernike Moments." International Conference on Pattern Recognition, Aug. 2004, Cambridge, UK. Accepted

Helose Hse, Michael Shilman, and A. Richard Newton. "Robust Sketched Symbol Fragmentation using Templates." International Conference on Intelligent User Interfaces, Jan. 2004, Madeira, Portugal. pp. 156-160 [pdf]

Copyright