中文摘要：

针对如何提取纸币图像特征的问题，提出了一种基于离散Haar小波变换和模糊逻辑相结合的纸币特征提取方法。该方法首先使用Haar小波对纸币图像进行分解操作，提取出图像的低频小波系数、高频小波系数。在此基础上引入模糊逻辑方法，把提取的小波系数分别作为语言变量，并构造出相应的隶属度函数，在模糊特征空间中求出每个模糊区域对应的激活强度值，将这些激活强度值进行归一化处理后构成纸币特征向量，使用神经网络分类器对纸币进行识别。此方法在资源约束的嵌入式系统（TITMS320C6713DSP）上实现，实验结果表明，离散Haar小波变换和模糊逻辑相结合的特征提取方法可以取得较高的识别率。

英文摘要：

A banknote feature extraction method based on the discrete Haar wavelet transform and the fuzzy logic is proposed aiming at extracting banknote feature efficiently. The Haar wavelet is applied to the operation of banknote image decomposition, which consequently results in the wavelet coefficients of low frequency and high frequency. Based on it, the theory of fuzzy logic is introduced to construct the corresponding membership function when the extracted wavelet coefficients are considered as linguistic variables respectively. The firing strength values of each corresponding fuzzy region are calculated in the fuzzy feature space, which then constitute the banknote feature vector after being normalized. The banknote recognition can be conducted by the neural network classifier. The above method was used to perform the experiment on the resource constrained embedded system of TI TMS320C 6713 DSP, and the results show that the high recognition rate was obtained.