Abstract: A new dimension reduction method was proposed based on two-dimension discrete wavelet decomposition and block discrete cosine transform. The new method was compared with the famous PCA-LDA method which was widely used for feature extraction and dimension reduction in pattern recognition. The experimental results show that the new method could get approximately equal retrieval accuracy to PCA-LDA method, but it performed better in calculation complexity and speed than the latter. So the proposed method is an effective dimension reduction means for ear recognition.

Key words: ear recognition, wavelet transform, Haar wavelet, discrete cosine transform, block DCT

摘要： 提出一种新的对人耳图像进行降维的方法。先对人耳图像进行二维的离散小波分解，然后再对小波分解后得到的低频信息施行分块离散余弦变换，进而获得图像的特征向量。实验证明，该方法与模式识别领域中广泛应用的PCA-LDA方法相比，在识别率大体相当的前提下，具有计算量小、降维速度快的优点，是对人耳图像进行特征提取的一种有效的手段。

关键词: 人耳识别, 小波变换, Haar小波, 离散余弦变换, 分块DCT