基于多分辨率的多层分类器的手语识别方法

A Method of Sign Language Recognition Based on Multiresolution Multilayer-Classifier

本文提出了一种基于多分辨率的多层分类器的手语识别方法,该方法对来自数据手套的手语输入,先用多分辨率选择特征,然后根据这些特征数据先进行低分辨率识别,再使用全部数据进行高分辨率识别。实验结果表明,该方法比传统HMM(隐马尔可夫模型)识别过程识别速度平均提高了约0.6秒,识别率提高了6.73％。

A method of sign language recognition based on multiresolution multilayer-classifier is proposed in this pa- per. For the input from data gloves, firstly features are selected using multiresolution theory, then low-resolution recognition with those selected features is made, at last high-resolution recognition with all the data is executed. The result of experiments shows that the recognizing average speed is 0.6 second faster than the traditional HMM recogni- tion process, and the recognizing ratio has been enhanced by 6.73%.