An algorithm of Handwritten Digital Recognition Based on BP-Bagging
Zuojun Liu, Lihong C. Li, Yu Mi · 2015
The handwritten number recognition algorithm based on BP-bagging generates the basic classification by BP network and generates multiple classifiers by bagging.The algorithm treats a handwritten character as an image.By scanning the images, 25 dimension features are extracted, and then compresses the 25 dimensions features into 5.The handwrite digital can be recognized by input it into BP-bagging classifier and by multiple voting integration.Proved by repeating experiments, the algorithm bears a high recognition rate, which functions better than single classifiers and other basic classifier combination algorithm. Introduction.The computer is widely used in all kinds of fields.Automation and intellectualization is becoming universal.The recognition technology of handwritten digital has been widely used in zip code recognition and banking business etc.The recognition technology has become an important subject in many fields.The handwritten digital is differing in thousands of way, such as deformation, translation and scale change.The number of the digital is very few only 0 to 9 and the stroke of the digital is very simple.The writing of each digital is differing in various ways.The people living in different area all are using digitals.The writing takes on obvious regional characteristic.So, it's hard to develop a universal digital recognition system.In many fields, people require high recognition rate.How to improve the recognition rate has become a problem to be solved urgently.Massive scale parallel processing, fault tolerance and learnability are characteristics of BP neural network.After training, the neural network can be used in handwriting digital recognition.The invariant features of handwriting digital can be extracted by the neural network.Due to the weakness of itself, the neural network classifier belongs to weak classifier.It is difficult to improve the recognition rate of pure neural network.So, bagging is introduced in this paper.The weak classifier can be transformed into stronger one by introducing bagging. Classifier Design of Bagging.