Related Experiment Videos
Recognition of Telugu characters using neural networks
M B Sukhaswami1, P Seetharamulu, A K Pujari
1Department of Computer and Information Science, University of Hyderabad, India.
International Journal of Neural Systems
|September 1, 1995
Summary
This study introduces a novel Multiple Neural Network Associative Memory (MNNAM) for recognizing Telugu characters, outperforming traditional methods. The artificial neural network approach effectively handles noisy and distorted printed and handwritten characters.
Area of Science:
- Computer Science
- Artificial Intelligence
- Pattern Recognition
Background:
- Traditional pattern recognition methods for Telugu character recognition have limitations, especially with noisy or distorted inputs.
- Previous attempts at Telugu character recognition using artificial neural networks (ANNs) have not been fully explored or optimized.
Purpose of the Study:
- To develop and evaluate an improved system for recognizing printed and handwritten Telugu characters.
- To address the limitations of existing methods by employing advanced artificial neural network techniques.
- To enhance the accuracy and robustness of Telugu character recognition in the presence of noise and variations.
Main Methods:
- Initially employed the Hopfield model of neural network as an associative memory for character recognition.
- Proposed a novel scheme, Multiple Neural Network Associative Memory (MNNAM), to overcome the capacity limitations of the Hopfield network by using parallel processing.
- Developed a detailed preprocessing scheme for digitized Telugu characters.
Main Results:
- Demonstrated the suitability of the Hopfield network for recognizing noisy printed and diverse handwritten Telugu characters.
- The proposed MNNAM scheme effectively overcomes storage capacity limitations by combining multiple neural networks.
- Satisfactory recognition rates were achieved using the proposed MNNAM strategy and various learning techniques.
Conclusions:
- The Multiple Neural Network Associative Memory (MNNAM) offers a robust and effective solution for Telugu character recognition.
- Artificial neural networks, particularly the proposed MNNAM, provide a significant improvement over conventional methods for handling noisy and distorted characters.
- The developed preprocessing and recognition strategy shows promise for practical applications in digital document processing.