基于PCA+LDA的特征融合的3D手写识别特征集取技术

3D Space Handwriting Recognition Feature Extraction Based on PCA and LDA Combination

文章针对基于3D加速度传感器的空间手写识别进行特征集取的研究,提出了一种基于PCA+LDA的特征融合进行分类识别的方法。首先从加速度数据中提取时域特征-旋转特征RF(RotationFeature),然后再利用FFT变换(Fast Fourier Transform)提取三维加速度的频域特征FFT,接着将时域特征RF和频域特征FFT进行特征融合,并利用PCA(Principal Component Analysis)+LDA(Linear Discriminate Analysis)组合进行降维,最后利用SVM(支持向量机)进行分类识别。实验结果显示,本文提出的方法可以有效提升3D手写识别系统的识别率。

In the research on the feature extraction for 3D space handwriting recognition based on 3D accelerometer,a new recognition method based on fusion feature which combines PCA algorithm and LDA algorithm is proposed.The method can be explained as follows： firstly,from accelerometer data we extract the timedomain feature-Rotation Feature（RF）;secondly,we uses FFT to extract the frequency-domain feature-FFT Feature of the accelerometer data;then the above two categories of features are fused together and the Principal Component Analysis（PCA） and Linear Discriminate Analysis（LDA） Combination is applied to reduce the dimension of the fusion feature.Finally,Supported Vector Machine（SVM） is used in the recognition of the 3D space handwriting characters.The experiment results show that the proposed method can significantly improve the performance of 3D space handwriting recognition systems.