Drawinput - Handwriting recognition input for Android

This a project to demonstrate handwriting recognition input. Key technologies used in this project:

1. RBF SVM from LIBSVM
2. Simple preprocessing
   • duplicate point removal
   • character data resampled to 12 points / char
   • normalization of data
3. Simple feature extraction
   • 4 features per point
   • normalized (x,y) of pen coordinates
   • sin(dx/dy) and cos(dx/dy) to estimate pen direction
4. SVM models trained with data from Unipen database
   • different models for numbers, small abc, big abc and special characters
   • 15 samples / char for numbers
   • 30 samples / char for abc and ABC models
   • X samples / char for special characters