A Robust Handwritten Numeral Recognition Using Hybrid Orthogonal Polynomials and Moments.

Area of Science:

• Computer Science
• Artificial Intelligence
• Pattern Recognition

Background:

• Numeral recognition is vital for optical character recognition and document understanding.
• Existing methods struggle with accuracy and speed, especially in noisy environments.
• Robust handwritten numeral recognition is crucial for real-world applications.

Purpose of the Study:

• To propose a novel, fast, and robust handwritten numeral recognition scheme.
• To enhance feature extraction efficiency and classification accuracy.
• To evaluate the method's performance in both clean and noisy environments.

Main Methods:

• Utilized Hybrid Orthogonal Polynomials for gradient and smoothed feature extraction.
• Employed the embedded image kernel technique to reduce feature extraction complexity.
• Used Support Vector Machine (SVM) for numeral classification.
• Evaluated on Roman, Arabic, and Devanagari numeral datasets.

Main Results:

• The proposed method achieved high recognition accuracy across all tested datasets.
• Demonstrated significant robustness against noise distortion.
• Outperformed Convolutional Neural Networks (CNNs) in noisy environments.
• Achieved comparable or superior accuracy to state-of-the-art methods.

Conclusions:

• The Hybrid Orthogonal Polynomials scheme offers a feasible and effective solution for handwritten numeral recognition.
• The method excels in practical, noisy environments, surpassing current benchmarks.
• This approach provides a robust alternative to existing numeral recognition techniques.