基于DT-CWT和PCA的唇部特征提取方法

摘要

针对单视觉唇读系统中唇部特征的提取问题,提出了基于双树复小波和PCA的唇部特征提取方法.利用双树复小波变换的近似平移不变性,通过将变换后的系数幅值重新排列,克服了感兴趣区域中唇部偏移的影响.与DCT相比.双树复小波变换既可反映输入信号的频域特性,又能反映其空间域特性,具有近似平移不变性.这些特性使得DT-CWT+PCA的方法优于DCT+PCA的特征提取方法.实验结果表明,该方法在识别率方面有较大的提高.%For the lip feature extraction of visual only lip reading system, a method is proposed based on dual tree complex wavelet transform to extract the visual feature of lip. Using the approximate shift invariance of DT-CWT, by rearrange the coeffients to overcoming the influence of the lip shift in the region of interest. Compared with DCT, DT-CWT can reflect the characters of the signal both in space domain and frequency domain, and it has the characters like approximate shift invariance and good directional selectivity. All these make DT-CWT+PCA performance better than the method based on DCT+PCA, the experimental results show that the recognition accuracy improved greatly.