Research On Design Of Handwriting System Based On Kinect Sensor

Character writing systems,as an important way of human-computer interaction(HCI),can transfer information from Human beings to computer accurately and efficiently.At present,most popular character writing systems still rely on extra devices,such as keyboards,tablet,touch screens and so on.These stiff modes restrict users to a narrow space,unable to satisfy people's requirements for the speediness,convenience and highly freedom of HCI.Combining on-line handwritten Chinese character recognition and computer vision technology,this thesis implements a handwriting system based on Kinect.Different from the traditional handwriting system,it has the advantages of non-contact,real-time response,the use of the scene more lenient,so has a strong versatility.Using this system,people can regard the fingertip as a pen and write in the air.The main work and contributions in this paper are as follows:A handwriting system based on Kinect has been designed.It is a non-contact virtual handwriting system composed of gesture recognition module and handwriting track recognition module.Realized a hybrid hand segmentation method based on depth model and skin color model,which can handle the technical problems of traditional segmentation methods,such as skin color confusion,hand-face coincidence,light changes,etc.,to achieve accurate segmentation of hand images.Realized a gesture recognition method based on Hu invariant moment?perimeter area ratio and binary image,defined the relevant gesture,extracted the feature vectors,and used SVM as the classifier,control the system by recognizing gesture commands.Locating fingertips by a dual-mode method,the method can effectively solve the invisible problem of the finger and adapt to the varied hand gesture.In this paper,a character recognition scheme based on DTW and stroke number is realized,and a handwritten Chinese character library is trained,so that the exact recognition of handwritten Chinese characters is achieved.Experiments show that the recognition rates of inputting Chinese characters,English letters and digits are all over 93% for the first four candidates with a speed of 20 fps.It can sufficiently meet the demand of daily use.