SHF: Small: Novel SW/HW Approximate Computing Methodologies with Case Studies on Biometric Security Systems

越来越多地部署生物识别系统以满足移动的、嵌入式和物联网设备的安全需求。该项目的目标是设计技术，使生物识别系统具有更快的响应时间，更少的能耗和成本，而不影响准确性和安全性。除了主要的研究目标外，该项目还包括一个全面的教育和推广计划，其中包括本科生的研究经验，跨学科教育课程，高中生的推广以及向行业的技术转移。为了实现该项目的目标，将开发新的近似计算技术，以简化生物识别系统的硬件，而很少或根本不影响准确性。该项目研究了使用布尔矩阵分解进行近似逻辑电路综合的新方法，从而在电路复杂性和结果质量之间实现平滑的权衡。该项目还研究了使用重复学习技术来导航潜在近似设计的空间，以确定最佳设计。最后，该提案调查了使用所提出的方法对端到端生物识别系统案例研究的影响，包括指纹识别，虹膜扫描和3D面部识别。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估而被认为值得支持。

Biometric systems are increasingly deployed to meet the security demand of mobile, embedded and internet-of-things devices. The goal of this project is to devise techniques that lead to biometric systems with faster response time, less energy consumption and cost without compromising the accuracy and security. In addition to the main research goal, this project includes a comprehensive education and outreach plan that includes research experience for undergraduate students, interdisciplinary educational courses, outreach to high-school students, and tech transfer to industry. To achieve the goal of this project, new approximate computing techniques will be developed to simplify the hardware of biometric systems with little or no compromise to accuracy. The project investigates new methods for approximate logic circuit synthesis using Boolean matrix factorization, which leads to smooth trade-off between circuit complexity and quality of results. The project also investigates the use of reinforcement-learning techniques to navigate the space of potential approximate designs to identify the best ones. Finally, the proposal investigates the impact of using the proposed methods on case studies of end-to-end biometric systems, including fingerprinting, iris scanning, and 3D facial recognition.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.