Critical Review on Off-Line Sinhala Handwriting Recognition
Download as PPTX, PDF2 likes281 views
The document discusses the development of off-line Sinhala handwriting recognition techniques, highlighting the unique characteristics of the Sinhala language. It outlines the processes involved in handwriting recognition such as preprocessing, segmentation, feature extraction, and recognition, utilizing methods like hidden Markov models and neural networks. The recognition accuracy achieved includes 94% with unique shapes and 75% with confusing shapes, showcasing the effectiveness of the employed methodologies.
Critical Review on Off-Line Sinhala Handwriting Recognition
- 1. Off-Line Sinhala Handwriting Recognition D.G.A.M.Wijayarathna (144193K) Faculty of Information Technology University of Moratuwa
- 2. ● Sinhala is majority language of Sri Lanka ● Curve shaped letters ● Left to right direction ● 16 vowels, 40 consonants, 2 semi-consonants and 13 consonant modifiers ● Written within three horizontal layers. Sinhala Language
- 3. ● Rapidly increasing and growing technology ● Scanned documents ● Many practical usage ● Use handwritten methods for official and unofficial works ● Most difficult task in character recognition Off-Line Recognition of Handwriting
- 4. Sinhala Handwriting Recognition Developed Using ● Feature Extraction ● Hidden Markov Models ● Neural Networks(ANN) ● Genetic Algorithm Common procedure 1. Preprocessing 2. Segmentation 3. Feature Extraction 4. Recognition
- 5. 1. Preprocessing ● Data collection ● Noise removing and Binarization ● Using the thresholding technique
- 6. ● Segment characters and their modifiers 1. Segment Text Lines 2. Segment Horizontal Characters 3. Identifying Overlapping Characters 4. Segment Overlapping Characters 2.Segmentation
- 8. ● To increase the classification accuracy ● Maximizing the recognition rate ● Used as inputs for training and recognition systems 3. Feature Extraction
- 9. ● Statistical techniques ● Artificial Neural Networks (ANNs) ● Template matching ● Structural techniques 4. Recognition
- 10. Sinhala Handwriting Recognition Statistical techniques ● Hidden Markov Models Genetic Algorithms ● Genetic Algorithm Based Alphabet Training Approach Artificial Neural Networks ● Fuzzy Neural Hybrid method ● Neural network based on backpropagation
- 11. Summary ● Feature extraction resulted 94% accuracy ● Hidden Markov Method shows 64% accuracy ● ANNs technique shows ○ 94% accuracy with unique shapes ○ 75% rate with confusing shapes.
- 12. THANK YOU!
Editor's Notes
- #6: Jasekara udawatta-- character conversion, scale correction, thinning Dharmalapa -- noise reduction, bizarization, skew correcion(angle ) Chamikara -- rgb to gray, noise removal, binarization, skeltanization, normalization(resolution)
- #7: Jayasekara--- vertical projection, horizontal projection Dharmapala--- vertical projection, horizontal projection, contours for ovellaping characters, Chamikara-- segmented in to six segments
- #9: Jayasekara udawatta---- grouping using 3 layers Dharmapala--- feature vectors, vertical projection profiles, horizontal projection,
- #11: Hewawitharana used HMM Jayasekara ---Genetic Algorithm Based Alphabet Training Approach Dharmapala-- neural network based on back propegation Chamikara-- fuzzy neural hybrid method