Statistical method-based evolvable character recognition system

A novel character recognition system is proposed in this paper. By using the virtual reconfigurable architecture-based evolvable hardware, a series of recognition systems are evolved. To improve the recognition accuracy of the proposed systems, a statistical pattern recognition-inspired methodology is introduced. The performance of the proposed method is evaluated on the recognition of characters with different levels of noise. The experimental results show that the proposed statistical pattern recognition-based scheme significantly outperforms the traditional approach in terms of character recognition accuracy. For 1-bit noise, the recognition accuracy is increased from 84.8% to 96.7%. For 5-bit noise, the proposed system achieves a recognition accuracy of 84%.