This project report covers the development of a computational 3D microscope, NoScope. Using tomographic and light field algorithms, we present a method to reconstruct 3D volumes of microscopic samples taken with a lensless sensor. Business and intellectual property strategies for commercializing NoScope are detailed in the first three sections. The remaining sections highlight the project’s technical accomplishments and methods.

关键词：3D显微镜;传感器;知识产权

查看摘要 索取原文[59页]

In order to develop provably safe human-in-the-loop systems, accurate and precise models of human behavior must be developed. Driving and  flying are examples of human-in-the-loop systems that both require models of human behavior and are safety critical. This report describes a new experimental setup for human-in-the- loop simulations. We focus on developing testbed for collecting driver data that allows us to collect realistic data, while maintaining safety and control of the environmental surroundings. A force feedback simulator with four axis motion has been setup for real-time driving experiments. The simulator will move to simulate the forces a driver feels while driving, which allows for a realistic experience for the driver. This setup allows for  flexibility and control for the researcher in a realistic simulation environment. Experiments concerning driver distraction can also be carried out safely in this test bed, in addition to multi-agent experiments.

关键词：驱动;仿真驾驶;实验

查看摘要 索取原文[34页]

Queueing networks are used to model complicated processing environments such as data centers, call centers, transportation networks, health systems, etc. A queueing network consists of multiple interconnected queues with some routing structure, and a set of servers that have diff erent and possibly overlapping capabilities in processing tasks (jobs) of di fferent queues. One of the most important challenges in designing processing systems is to come up with a low-complexity and efficient scheduling policy.

关键词：排队网络;调度策略;路由结构

查看摘要 索取原文[102页]

Brain-machine interfaces (BMIs) aim to assist patients suffering from neurological injuries and disease by enabling them to use their own neural activity to control external devices such as computer cursors or robotic arms, or even drive movements of their own body via muscle stimulation. At the heart of a BMI system is the decoding algorithm, or “decoder”, that translates recorded neural activity into control signals for a prosthetic device. Decoders are often initialized offline by first recording neural activity while a subject performs real movements, or observes or imagines movements, and then fitting a decoder to predict these movements from the neural activity. However, BMIs are fundamentally closed-loop systems, since BMI users receive performance feedback (e.g. by visual observation of the prosthetic’s movements), and the prediction power of decoders trained offline does not directly correlate with closed-loop performance. In other words, a high level of BMI performance can not necessarily be achieved solely by optimizing decoder parameters in an open-loop setting.

关键词：脑机接口；体质指数；解码算法；闭环系统

查看摘要 索取原文[124页]

Analysis in the domain of Online Video Data Analytics. In the co-written portions of this document, we discuss the projected commercialization success of our products by analyzing worldwide trends in online video, presenting a competitive business strategy, and describing several approaches towards the management of our intellectual property. In the individually written portion of this document, we discuss our implementation of two machine learning models, k Nearest Neighbors and Random Forest, and evaluate them as a means of identifying subscription churners, the primary goal of Subscriber Analysis. In particular, we show that the performances achieved by these models using our initial, restricted feature set are promising and warrant future exploration of these models. 

关键词：网络视频；数据分析；机器学习模型

查看摘要 索取原文[71页]

This paper illustrates the design and implementation of a robotic software system based on robot-human interactions, which increases the applications of low-cost, inaccurate robots such as Baxter. The first four parts of the paper discuss the motivation, market & industry, IP strategy and system workflow. Then the paper explains a specific technical contribution of the project in detail – estimating Baxter’s positioning accuracy. Based on forward kinematics and Jacobian matrix respectively, two models are built to convert the error tolerance of the joints to the error tolerance of the end effector in the Cartesian space. The second model based on Jacobian matrix is chosen as the final approach since it has linear complexity and it is more computationally efficient. The result gives a reasonable estimation of Baxter’s accuracy, which allows valid comparison with the task’s accuracy requirements. Thus we are able to determine how the human can interact with the robot and improve its performance.

关键词：机器人；人工交互；雅可比矩阵；机器人软件系统

查看摘要 索取原文[38页]

For the past century, photographs have served as reliable primary sources of evidence, but that is quickly changing. Photo manipulation tools have become widespread and it is easy to manipulate images. Photo manipulations tools such as Adobe PhotoShop afford greater artistic expression, and enable users to create manipulations that challenge the limits of our natural perception. The difference between authentic and manipulated photos has become harder to distinguish, and can only be detected by digital forensic experts. The image forensics capstone project aims to create an online software service that performs the work of forensic analysts, and visualizes and analyzes the possible manipulations that may have been performed on an image. Our goal is to empower anybody to perform image forensic analysis, and help increase our faith in digital content.

关键词：图像处理；图像识别；可视化分析

查看摘要 索取原文[45页]

Speech is possibly the most efficient way humans can communicate with a machine. If acoustic waves could perfectly map to words, a variety of exciting applications could be realized. However, speech recognition is difficult in reality. One reason is that there is much variability in people’s voices. In addition, many applications of speech-recognition require voice authentication or speaker-recognition in order to be practically useful. Speaker identification systems can account for voice variability, and also can provide voice authentication, and hence are an interesting area of research.

关键词：通信;语音识别;声音认证

查看摘要 索取原文[45页]

Low power wireless medical electronic design is a heated research topic due to the development of remote medical diagnosis systems and long-term treatment assistances. In general, such medical devices contain bio-signal sensor that monitor the health status of a patient or health-conscious person. Our main goal is to design an LNA that is suitable for amplifying bio-signals in low-power remote sensing devices. In this work, the business strategy and the IP strategy is being studied and presented.

关键词：低功耗无线医疗；环扩增技术；生物信号LNA

查看摘要 索取原文[52页]

Online social networking services (e.g., Facebook, Twitter, and Blogger) bring new bene fits to almost all aspects of our lives. They have completely transformed how we communicate with each other, how we process information, and how we diff use social influence. However, these social networking services are also plagued by both conventional and emerging threats to security and privacy. For instance, two fundamental security risks are 1) users' accounts are compromised by attackers or get lost and 2) attackers create massive fake (or Sybil) accounts to launch various malicious activities. In this thesis, we rst design secure and usable account recovery methods based on users' trusted friends to recover compromised or lost user accounts. Second, we construct a scalable semi-supervised learning framework, which is based on probabilistic graphical model techniques, to detect Sybil accounts. Third, we demonstrate that diverse private information (e.g., private user demographics and hidden social connections) can be inferred with high accuracies from data that is publicly available on social networking sites, which has implications for the design of privacy-preserving online social networking services.

关键词：在线社交网络服务；隐私保护；安全性；学习框架

查看摘要 索取原文[97页]