基于手掌电阻抗谱多频率同步测量的生物特征识别新方法

Biometrics has been considered as the most potential identity authentication technology. But the current biometrics technologies, which are generally based on 2D or 3D image feature, could not reach a balance among anti-falsification, usability, non-invasiveness and cost, so that their popularization is still restricted. This project points out that the bio-impedance spectroscopy (BIS) of human palm is a new biometric feature. The researches in this project include: Firstly, the Maximum Length Sequence (MLS) is adopted as the excitation signal, and a new multi-frequency synchronized measurement of BIS method based on the principle of system correlation identification is proposed, which could realize BIS measurements at mass frequency points in a short time. Secondly, the fast algorithm will be studied, and a new BIS measurement system will be established based on the proposed method, through which, large number of palm BIS sample data will be acquired from many volunteers. Thirdly, the Immune Clonal Selection Algorithm is adopted to achieve the optimal feature subset, and a classifier based on Hidden Space Supporting Vector Machine Algorithm will be designed, whose performance will be evaluated by the Bootstrap. Ultimately, the biometric recognition mechanism based on palm BIS measurements will be established, which will settle the theoretical and technological foundation for further development of corresponding biometrics device. This novel biometrics method based on the impedance spectroscopy (non-image) feature has the outstanding advantage of high anti-falsification, flexible usability, non-invasiveness and low-cost, and will be hopeful to become an important complement of the current biometric technologies, which indicates that this project have significant research value.

生物特征识别是最具潜力的身份认证技术，但当前生物特征识别技术大多基于二维或三维的图像特征，在防伪性、无创性、便捷性和低成本之间不能兼顾，普及性仍受制约。本项目首次提出手掌的电阻抗谱(BIS)特性是一种新的生物特征，研究包括：采用最大长度序列作为激励源，提出基于系统相关辨识原理的BIS多频率同步快速测量方法；研究快速测量算法，构建手掌BIS测量系统实现海量频率点阻抗的快速测量；以实验获得的手掌BIS数据作为原始特征，研究基于免疫克隆算法的特征选择方法获得原始特征的最优特征子集，设计基于隐空间支持向量机的分类器，并通过自助抽样法评估分类器性能，最终建立基于手掌BIS测量的生物特征识别新方法，为进一步研制相应的生物特征识别设备奠定理论和技术基础。这种基于电阻抗谱(非图像)特征的生物特征识别方法具有防伪性强、无创、便捷、廉价等突出优点，有望成为现有生物特征识别方法的重要补充，具有重要的研究价值。

生物特征识别是最具潜力的身份认证技术，开展原创性的、具有自主知识产权的生物特征识别技术，对国家公共安全、信息安全具有深远的意义。本项目提出一种全新的基于手掌电阻抗谱(BIS)测量的生物特征识别新方法，为生物特征识别的研究注入了新思路，具有重要的学术研究价值和可预期的应用前景。主要研究内容包括：①利用最大长度序列（MLS）的伪随机特性，提出一种基于系统相关辨识原理的BIS多频率同步快速测量方法，并发现在MLS信号一个码元内采样多次（过采样）可以显著提高BIS的测量精度；②设计了基于ChirpZ变换的频谱分析算法，大大减少了由于过采样而引起的大计算量；③以FPGA+ARM为核心，设计了基于系统相关辨识原理的BIS测量系统，1ms时间内可以完成1023个频率点的测量，实现了BIS的快速测量；④提出一种改进的低波峰因数（CF）的Multisine合成算法，四种不同的频率分布和幅值分布的Multisine信号合成实验证明：该算法合成的Multisine信号具有最小的CF值；⑤在扫频式手掌BIS测量系统研制的基础上，提出一种全新的手指感应容抗谱（CRS）测量方法，并研制了手指感应CRS测量仪，实现了4~1024kHz频率范围的256的点CRS测量，并设计出了较为理想的测量电极，选择出了较为理想的信号线类型与尺寸；⑥利用新研制的手指感应容抗谱(CRS)测量仪进行了系统的试验研究，对500名志愿者进行了手指感应CRS测量实验，分别使用SVM和BP神经网络算法进行了BR实验研究，识别率分别达到81%和85%。上述研究结果初步证明了基于手指感应CRS测量的生物特征识别方法具有一定的可行性。项目执行期间累计发表论文13篇（其中SCI论文8篇），申请/获得国家发明专利5项，软件著作权3项，累计培养硕士研究生15名。