A handwritten digit recognition method based on hybrid features extraction and its application to a temperature sensor array

This paper proposed a hybrid feature extraction method to improve the correct recognition rate of a handwritten digit recognition device based on temperature sensor array. The hybrid features are based on the temperature changes of the temperature sensor array during the process of handwriting, and the Principal Component Analysis (PCA) method is used for choosing the principal component of the features. Then the Support Vector Machine with the kernel of Radial Basis Function (RBF) is used for the online handwritten digit recognition. Lastly the above methods are applied in the online handwritten digit recognition system based on the temperature sensor array, and its performance is evaluated with well-designed comparative experiments. The experimental results demonstrate that the correct recognition rate of the hybrid features extraction based method exceeds 99%, which is 4% better than the static features based method and 37.5% better over the dynamic features based method.