SBIR Phase I: Matching Filter Solution for High Accuracy Signature Dynamics Verification

这个小型企业创新研究第一阶段研究项目开发了一种先进的、高精度的签名验证系统，该系统基于一种新的笔迹动力学信号处理二维波形-笔的加速度的x和y坐标。研究的关键创新在于开发了特殊的信号处理技术，将非平稳的动态手写信号转换为平稳的手写动态信号。这促进了匹配过滤器解决方案，并随后克服了签名验证技术精度不足的长期瓶颈，这阻碍了其在更广泛的市场中的应用。该方法实现了用于签名验证的匹配滤波器，在理论上是评估信号对统计相似性的最优方法。现有技术的核心问题在于它们依赖于个人签名的平均特征作为识别的模板。然而，这些平均特性具有显著的离散性，这降低了系统的精度。由于手写动态信号的非平稳性，以及手写过程中随机的毫秒级时延，匹配滤波概念在签名验证中的应用长期以来一直受到阻碍。拟议项目的主要目的是为安全和隐私保护创建一种方便、可靠的生物特征识别器。该装置具有很大的潜在商业价值。广泛的客户类别包括：联邦调查局、中央情报局、其他执法机构，用于网络安全，以及用于保护对敏感和受限区域的物理访问的安全，如核电站和化工厂；护照控制；医疗提供者和保险公司，用于保护医疗记录。商业部门的应用包括金融交易的安全，例如电子资金转账、在线银行和其他金融交易，以及进入自动柜员机的卡访问。

This Small Business Innovation Research Phase I research project develops an advanced, high-accuracy signature verification system based on a novel handwriting dynamics signal processing of two-dimensional waveform - the x- and y-coordinates of the pen's acceleration. The key innovation of the research is the development of special signal processing techniques that convert the non-stationary, dynamic handwriting dynamic signal to a stationary one. This facilitates a matched-filter solution, and subsequently overcomes the long-standing bottleneck of insufficient accuracy of signature verification techniques, which has forestalled its application to broader markets. The proposed approach makes possible the implementation of a matched- filter for signature verification, which is the theoretically optimal method for the evaluation of statistical similarity of a signal pair. The core problem with existing techniques lies in their reliance on the mean characteristics of an individual's signatures as a template for identification. However, these mean characteristics have significant dispersion that degrades the accuracy of the systems. The application of matched filtering concept to signature verification has been impeded for long time due to the handwriting dynamic signal's non-stationarity, resulting from random milliseconds-long time delays during the handwriting process. The main purpose of the proposed project is to create a convenient reliable biometric identifier for security and privacy protection. The proposed device has a great potential commercial value. The broad customer categories include: Federal Bureau Investigations, Central Intelligence Agency, other law enforcement agencies for cyber security, as well as for security of physical access to sensitive and restricted areas, such as nuclear and chemical plants; passport control; medical providers and insurance companies for protection of medical records. The commercial sector applications include security of financial transactions, e.g. electronic fund transfers, on-line banking and other financial transactions, and card access to automated teller machines.