Biometric Intelligent Interfaces

拟议的研究计划设想为环境生物识别传感和识别环境建立一个统一的框架。该框架将集成新兴的准确和可感知的潜在生物识别、智能图像/信号处理和高级决策支持。 该框架将产生控制界面，提供对手势、面部和面部表情以及身体的无缝扫描，以识别和持续跟踪用户的行为。这将通过加强生物识别技术的使用，将自然、直观和沉浸式(NII)界面的概念与NIIS访问的安全性结合起来。这不仅允许在安全访问控制应用中集成多个NII，而且还将加强相关NII应用的安全性和隐私，例如用于个人设备、计算机系统和态势感知环境的非接触式接口。 该框架的中间件将建立在使用应用编程接口的平台上，并使用图形处理单元执行低功耗的本地处理以及与高性能的应用处理并行的传感器处理。 我们将通过将生物识别传感器与智能上下文感知支持相结合来解决这一技术问题。这将涉及使用基于内容的多模式数据分析和索引来存储和处理多模式数据。在本研究计划中，将通过对多模式输入和多媒体数据的句法和语义分析来解决这一问题，以便提供智能或上下文感知的存储多媒体检索。 另一个挑战是低功率/低实施成本和低传感器精度之间的权衡。解决这个问题的方法包括使用多传感器方法，以及在传感器中嵌入更多的智能以补偿质量。这项提议将部署最先进的RGB深度传感器，例如目前安装在PrimeSense的Carmine中的多传感器方法，微软的Kinect可以感知和识别物体，使用0.8-3米距离物体的深度数据来感知和识别面部和全身运动，以及Leap 3D运动控制，它可以在0.2-0.8米处以卓越的精度(1/100 mm)识别用户的手指手势。 该项目将在卡尔加里大学生物识别技术实验室实施，以以前和目前的项目为基础，涉及指纹、虹膜和人脸、面部表情识别、可见光和红外光谱中的人脸识别等生物识别建模，以及根据多传感器和多模式结构和贝叶斯决策为新一代态势感知生物识别系统创建原型模块。 这项研究计划的最终愿景是将这个生物识别平台提升到NII的水平，这将影响底层技术的所有级别，实现预期的应用(在人与个人设备交互、安全、医疗保健和教育方面)，并创建一系列等待发现的未来应用。

The proposed research program envisions building a unified framework for an ambient biometric sensing and identification environment. The framework will integrate emerging accurate and perceivably latent biometric sensing, intelligent image/signal processing, and advanced decision-making support. The framework will result in control interfaces that provide seamless scans of gestures, face and facial expression, and body, for identification and continuous tracking of a user's actions. This will enable the combination of the concept of Natural, Intuitive and Immersive (NII) interfaces, and the security of the access to the NIIs, through enhanced usage of biometric technologies. This will not only permit integration of multiple NII in security access control applications, but will also reinforce security and privacy of related NII applications, such as contactless interfaces for personal devices, computer systems, and situational awareness environments. The middleware of the framework will be built upon a platform that uses an Application Programming Interface, and performs local processing for low power, as well as sensor processing in parallel with application processing for high performance, using Graphic Processing Units. We will address this technology issues by integration of biometric sensors with intelligent context-aware support. This will involve multimodal data storage and processing using content-based multimodal data analysis and indexing. This will be addressed in this research program via syntactic and semantic analysis of multimodal input and multimedia data, in order to provide intelligent, or context-aware, retrieval of stored multimedia. The other challenge is a trade-off between low power/cost of implementation and low precision of the sensors. Ways to resolve this problem include using multi-sensor approach, as well as embedding more intelligence in sensors to compensate for quality. This proposal will deploy the most advanced RGB-Depth sensors, such as the multi-sensor approach currently embodied in PrimeSense’s Carmine, Microsoft's Kinect that perceives and identifies objects, face and full body motions using depth data from objects at 0.8 -3m distance, as well as the Leap 3D Motion control that allow recognition of users' finger gestures with superior precision (1/100 mm) at 0.2-0.8 m. The project will be implemented in the Biometric Technologies Laboratory at the University of Calgary, and is based on the previous and current projects, involving modeling of biometrics such as fingerprint, iris and face, facial expression recognition, face recognition in both visual and infrared spectra, as well as creating the prototyping modules for a new-generation of situational awareness biometric systems, based on multi-sensor and multi-modal structure and Bayesian decision-making. The ultimate vision of this research program is to bring this biometric platform to the level of NII that will influence all levels of the underlying technology, enable the anticipated applications (in human-personal device interaction, security, health care and education), and create a whole host of future applications waiting to be discovered.