Research Of Preprocess Technique For 3D Space Handwriting Based On 3D Accelerometer

Space handwriting as a new technology for human-computer interaction is becoming the trend of development of the handwriting field, relying on its flexibility and comfort of the user experience. With the development of MEMS technology the input device of space handwriting is generally MEMS acceleration sensor which is small, high precision and has strong resistance to the external environment interference. In the current study of the space handwriting recognition field, the existence of the connected stroke seriously affects the recognition efficiency. In addition, the study of the space handwriting is in the initial stage, so there is not a mature character library, the data is insufficient when the classifier trains sample.In this study, the characteristics of the space handwriting characters were analyzed and a preprocess method was proposed which can eliminate the connected stroke and effectively solve the above problems. In this method, first using principal component analysis (PCA) method space characters were flattened, and then the characteristics of turning point and the characteristics of the stroke direction were extracted. In order to avoid the mistake in the elimination of connected stroke, support vector machine (SVM) was used to divide unknown characters into two categories, with or without connected stroke. Pretreatment approach not only eliminates the impact of stroke character recognition, but also transforms the space character trajectory to plane character trajectories. Finally, the flat character classification and the existing flat character library can be used to identify characters, which can effectively solve the problem of lack of data when the classifier trains sample.Finally, experiments were designed to verify the validity of the flattened of space characters based on principal component analysis method. The effect of the elimination of connected stroke of space handwritten characters after character classification based on SVM was showed. In addition, the HMM (Hidden Markov Model) was used as a classifier to recognize the character, the experiments showed that the recognition rate of characters has improved significantly after elimination of connected stroke, which verify the effectiveness of the preprocess method.