Integrated approach to handwritten character recognition using ANN and it´s implementation on ARM

Offline recognition is preferred when user tends to write a character in many different ways. Whereas, in online recognition, the way user writes, that is hand movements, are tracked. The difficulties like differentiating similar characters and hand movement dependence arises when these methods are applied individually. In order to overcome these problems, we propose Integrated Offline-Online Character Recognition method in specific manner to obtain optimum results out of them. For offline method, feature vector is found by vertical and horizontal scanning method. For online method, it is found by x-y concatenation method. Then they are fed to Feed-Forward Back Propagation Artificial Neural Networks in each method and final results are obtained by averaging. This algorithm is computationally efficient. This algorithm was successfully implemented and verified on ARM7-core based controller. So system has advantages like small size, low power consumption and low cost. We obtained accuracy of 99.12% for proposed method.