Michigan State University

MSU Libraries

Digital Repository

Search results

(1 - 14 of 14)
Title
Natural language based control and programming of robotic behaviors
Creator
Cheng, Yu (Graduate of Michigan State University)
Date
2019
Collection
Electronic Theses & Dissertations
Description
"Robots have been transforming our daily lives by moving from controlled industrial lines to unstructured and dynamic environments such as home, offices, or outdoors working closely with human co-workers. Accordingly, there is an emerging and urgent need for human users to communicate with robots through natural language (NL) due to its convenience and expressibility, especially for the technically untrained people. Nevertheless, two fundamental problems remain unsolved for robots to working...
Show more"Robots have been transforming our daily lives by moving from controlled industrial lines to unstructured and dynamic environments such as home, offices, or outdoors working closely with human co-workers. Accordingly, there is an emerging and urgent need for human users to communicate with robots through natural language (NL) due to its convenience and expressibility, especially for the technically untrained people. Nevertheless, two fundamental problems remain unsolved for robots to working in such environments. On one hand, how to control robot behaviors in dynamic environments due to presence of people is still a daunting task. On the other hand, robot skills are usually preprogrammed while an application scenario may require a robot to perform new tasks. How to program a new skill to robots using NL on the fly also requires tremendous efforts. This dissertation tries to tackle these two problems in the framework of supervisory control. On the control aspect, it will be shown ideas drawn from dynamic discrete event systems can be used to model environmental dynamics and guarantee safety and stability of robot behaviors. Specifically, the procedures to build robot behavioral model and the criteria for model property checking will be presented. As there are enormous utterances in language with different abstraction level, a hierarchical framework is proposed to handle tasks lying in different logic depth. Behavior consistency and stability under hierarchy are discussed. On the programming aspect, a novel online programming via NL approach that formulate the problem in state space is presented. This method can be implemented on the fly without terminating the robot implementation. The advantage of such a method is that there is no need to laboriously labeling data for skill training, which is required by traditional offline training methods. In addition, integrated with the developed control framework, the newly programmed skills can also be applied to dynamic environments. In addition to the developed robot control approach that translates language instructions into symbolic representations to guide robot behaviors, a novel approach to transform NL instructions into scene representation is presented for robot behaviors guidance, such as robotic drawing, painting, etc. Instead of using a local object library or direct text-to-pixel mappings, the proposed approach utilizes knowledge retrieved from Internet image search engines, which helps to generate diverse and creative scenes. The proposed approach allows interactive tuning of the synthesized scene via NL. This helps to generate more complex and semantically meaningful scenes, and to correct training errors or bias. The success of robot behavior control and programming relies on correct estimation of task implementation status, which is comprised of robotic status and environmental status. Besides vision information to estimate environmental status, tactile information is heavily used to estimate robotic status. In this dissertation, correlation based approaches have been developed to detect slippage occurrence and slipping velocity, which provide grasp status to the high symbolic level and are used to control grasp force at lower continuous level. The proposed approaches can be used with different sensor signal type and are not limited to customized designs. The proposed NL based robot control and programming approaches in this dissertation can be applied to other robotic applications, and help to pave the way for flexible and safe human-robot collaboration."--Pages ii-iii.
Show less
Title
Teleoperation of mobile manipulators
Creator
Jia, Yunyi
Date
2014
Collection
Electronic Theses & Dissertations
Description
Mobile manipulators provide larger working spaces and more flexibility than standard manipulators by introducing mobility. Through teleoperation, they can be applied to a variety of areas such as hazardous material handling, outer space exploration, searching and rescue, etc.Inspired by application requirements, there are four major challenges in the teleoperation of mobile manipulators including the modeling and control of mobile manipulators, teleoperation of multiple mobile manipulators,...
Show moreMobile manipulators provide larger working spaces and more flexibility than standard manipulators by introducing mobility. Through teleoperation, they can be applied to a variety of areas such as hazardous material handling, outer space exploration, searching and rescue, etc.Inspired by application requirements, there are four major challenges in the teleoperation of mobile manipulators including the modeling and control of mobile manipulators, teleoperation of multiple mobile manipulators, modeling the human teleoperator in teleoperation system and communications between the human teleoperator and mobile manipulators. Therefore, this study aims to address these challenges.For the modeling and control of mobile manipulators, the motion accuracy of the end-effector is a problem for the existing methods due to the system performance differences. To address this issue, we introduce a new control method with online motion distribution and coordination to improve the accuracy. In addition, a sensor-based redundancy resolution scheme is proposed to further improve the teleoperation efficiency.For the teleoperation of multiple mobile manipulators, the system stability under random communication delays and unexpected events is a major problem for the existing methods. To address this issue, we propose a non-time based teleoperation and coordination method. A non-time perceptive reference is designed as the new reference to replace the time in the system modeling and control. Through this design, the system stability under random communication delays and unexpected events could be ensured.For modeling the human teleoperator in teleoperation system, there are no existing models and the teleoperation efficiency and safety are always subject to the operation status of the teleoperator. To address this issue, we propose a concept named quality of teleoperator (QoT) to represent the teleoperator and incorporate it into the modeling and control of the teleoperation system. Through this design, the teleoperation efficiency and safety could be improved under various operation status of the teleoperator.For the communications between the human teleoperator and mobile manipulators, the existing methods of using joysticks are neither efficient nor intuitive. Therefore, we introduce the natural language as a new communication manner. However, the existing natural language control methods could not online handle unexpected events in the environment and robotic system. To address this issue, a new systematic natural language modeling and control method is designed to online handle such unexpected events.Finally, the proposed methods are all implemented on our developed mobile manipulators and the experimental results illustrate their effectiveness and advantages.
Show less
Title
Optimizing message to virtual link assignment in Avionics Full-Duplex Switched Ethernet networks
Creator
Klonowski, Joseph
Date
2019
Collection
Electronic Theses & Dissertations
Description
"Avionics Full-Duplex Switched Ethernet (AFDX) is an Ethernet-based data network that provides deterministic performance, high reliability, and lower costs and development time by utilizing commercial off-the-shelf networking components. As AFDX networks have become of the network are continually being evaluated. There are two main types of solutions to improving network performance: changes to the physical layer and changes to the logical layer. Because the physical network is setup prior to...
Show more"Avionics Full-Duplex Switched Ethernet (AFDX) is an Ethernet-based data network that provides deterministic performance, high reliability, and lower costs and development time by utilizing commercial off-the-shelf networking components. As AFDX networks have become of the network are continually being evaluated. There are two main types of solutions to improving network performance: changes to the physical layer and changes to the logical layer. Because the physical network is setup prior to defining the data that is transferred on the network, logical layer optimization becomes important and is often the only viable solution. Previous research has explored optimization of different aspects of the logical solution for a given target (whether it be latency or bandwidth), however, an approach for a customizable target using optimization techniques has not been attempted. In this work, we provide an overview of AFDX networks and discuss factors engineers consider while optimizing the network. Previously researched solutions are evaluated for effectiveness. We identify the need for an optimization solution that allows for a customizable objective to account for both message latency and bandwidth. To fill this gap, we consider the problem of assigning messages to virtual links, which are configurable, logical unidirectional links from publishing end systems to one or more subscribing end systems. We propose a flexible framework based on particle swarm optimization (PSO) that performs message to virtual link assignment in AFDX networks to optimize a user-defined objective. We discuss and provide results on PSO optimization for a range of hyperparameters. Finally, results for a sample swarm are presented to prove the feasibility and usefulness of the proposed approach."--Page ii.
Show less
Title
Fairness in AI-based recruitment and career pathway optimization
Creator
Mujtaba, Dena Freshta
Date
2020
Collection
Electronic Theses & Dissertations
Description
Work has long been a source of human livelihood, financial security, mental and physical well-being, dignity, and meaning. However, advances in computing, big data, artificial intelligence (AI), robotics, and related technologies are expected to usher in unprecedented and widespread changes in the economy and society. It is estimated that by 2030 up to 14% of the global workforce may need to change occupational categories as the world of work is disrupted by technological advances. Many...
Show moreWork has long been a source of human livelihood, financial security, mental and physical well-being, dignity, and meaning. However, advances in computing, big data, artificial intelligence (AI), robotics, and related technologies are expected to usher in unprecedented and widespread changes in the economy and society. It is estimated that by 2030 up to 14% of the global workforce may need to change occupational categories as the world of work is disrupted by technological advances. Many current and future workers that will enter the workforce lack skills that in-demand and future jobs require. In short, the landscape of work is poised for a major and unprecedentedly rapid transformation and this calls for a variety of strategies to meet the needs of workers, employers, the economy, and broader society.Motivated by these concerns, we investigate two key problems faced by organizations and workers in the future of work. As AI has expanded into human resource applications, organizations are increasingly using AI-based recruitment for sourcing, screening, and selecting talent. We explain how this can lead to biases in decisions and how this bias can be measured, review tools available for bias mitigation, and discuss future challenges for fairness in machine learning specific to recruitment applications. Alongside this, workers are affected not only by biased recruitment, but by the growing automation of tasks in occupations, which will increasingly require job and task transitions. To help workers navigate these transitions effectively, we propose a genetic-algorithm-based optimization engine to search for a worker's optimal career pathway in a network of occupations, given their current knowledge, skills, abilities, and other work-related characteristics. Overall, this thesis presents strategies for organizations to mitigate bias in AI-based recruitment and for workers to plan their career pathway in the face of unprecedented changes in the world of work.
Show less
Title
A neural recording front end for multi-channel wireless implantable applications
Creator
Li, Haitao
Date
2011
Collection
Electronic Theses & Dissertations
Description
A great demand exists for technologies that enable neuroscientists and clinicians to simultaneously observe the activity of many neurons in the brain. By recording this activity, awake animal behavior can be predicted in real time, brain-machine interfaces controlling the machine by thought can be set up, and treatments for neurological disorders can be explored. Existing commercial neural recording equipment are bench-top systems that are bulky, high cost, and consume high power. They also...
Show moreA great demand exists for technologies that enable neuroscientists and clinicians to simultaneously observe the activity of many neurons in the brain. By recording this activity, awake animal behavior can be predicted in real time, brain-machine interfaces controlling the machine by thought can be set up, and treatments for neurological disorders can be explored. Existing commercial neural recording equipment are bench-top systems that are bulky, high cost, and consume high power. They also require wire bundles tethering the neural recording probes to skull-mounted a connector that lead to tissue infection, external noise and interfering signals coupling. To overcome these disadvantages, a miniature wireless implanted multi-channel integrated neural recording micro-system with low power and low noise is needed. This thesis contributes to the analog front end of such a micro-system, which provides a low-power, low-noise neural interface that detects and amplifies neural signals and digitizes them for further signal processing. The front end includes neural amplifiers and an analog-to-digital convertor (ADC). This thesis work addresses the challenges to developing an analog front end for wireless implanted multi-channel neural recording systems, which include ultra low noise, extremely low power, high power supply rejection radio, low area occupation, sufficient data conversion speed and optimizing design tradeoff between all these requirements. Two versions of a neural amplifier were built. The second version was optimized based on the design experience of the first version and a comprehensive theoretical analysis of neural amplifiers. Following the optimization guidelines, noise efficiency and a new figure of merit for neural amplifiers were effectively improved. A successive approximation (SAR) ADC tailored to wireless implantable neural recording systems was also designed. The new SARADC is able to process 32 neural spikes recording channels in a multiplexing manner with low power consumption and low area occupation. The results of this research lay a solid foundation for future realization of high sensitivity wireless implantable neural recording system.
Show less
Title
Infrared sensing systems using carbon nanotube based photodetectors
Creator
Chen, Hongzhi
Date
2012
Collection
Electronic Theses & Dissertations
Description
Infrared (IR) sensing systems have versatile applications; however, low performance and high cost of the conventional photodetectors have prevented their widespread utilization in various fields. Carbon nanotube (CNT), a promising nano-material with excellent electrical and optical properties, has potential to develop high performance IR detectors compared to their conventional counterparts. However, there are three major difficulties that impede the application of CNT based photodetectors...
Show moreInfrared (IR) sensing systems have versatile applications; however, low performance and high cost of the conventional photodetectors have prevented their widespread utilization in various fields. Carbon nanotube (CNT), a promising nano-material with excellent electrical and optical properties, has potential to develop high performance IR detectors compared to their conventional counterparts. However, there are three major difficulties that impede the application of CNT based photodetectors for imaging systems. Firstly, there are challenges in design, fabrication, and testing of CNT photodetectors because of their nano-scale size and unique geometry. Secondly, the small diameter of the CNTs results in low fill-factor (absorption area). Thirdly, it is difficult to fabricate large scale of photodetector array for high resolution focal plane due to the limitations on the efficiency and cost of the manufacturing. The issues related to the design and fabrication of CNT based photodetectors were addressed by configuring the device in field effect transistors with different metal and gate structures. In addition, the theoretical foundations as well as the implementation schemes for the development of nano-structure lens to improve absorption efficiency of IR detectors were developed. The topics include the optical antennas that confine light into a sub-wavelength volume, as well as photonic crystals to increase the effective absorption area. Furthermore, a novel CNT based IR sensing system was developed. The experimental results showed that the new IR sensing system can achieve the superb performance enabled by CNT based photodetectors, and, at same time, to obtain high resolution and efficient imaging.
Show less
Title
Development of plenoptic infrared camera using low dimensional material based photodetectors
Creator
Chen, Liangliang
Date
2016
Collection
Electronic Theses & Dissertations
Description
Infrared (IR) sensor has extended imaging from submicron visible spectrum to tens of microns wavelength, which has been widely used for military and civilian application. The conventional bulk semiconductor materials based IR cameras suffer from low frame rate, low resolution, temperature dependent and highly cost, while the unusual Carbon Nanotube (CNT), low dimensional material based nanotechnology has been made much progress in research and industry. The unique properties of CNT lead to...
Show moreInfrared (IR) sensor has extended imaging from submicron visible spectrum to tens of microns wavelength, which has been widely used for military and civilian application. The conventional bulk semiconductor materials based IR cameras suffer from low frame rate, low resolution, temperature dependent and highly cost, while the unusual Carbon Nanotube (CNT), low dimensional material based nanotechnology has been made much progress in research and industry. The unique properties of CNT lead to investigate CNT based IR photodetectors and imaging system, resolvingthe sensitivity, speed and cooling difficulties in state of the art IR imagings.The reliability and stability is critical to the transition from nano science to nano engineering especially for infrared sensing. It is not only for the fundamental understanding of CNT photoresponse induced processes, but also for the development of a novel infrared sensitive material with unique optical and electrical features. In the proposed research, the sandwich-structured sensor was fabricated within two polymer layers. The substrate polyimide provided sensor with isolation to background noise, and top parylene packing blocked humid environmental factors. At the same time, the fabrication process was optimized by real time electrical detection dielectrophoresis andmultiple annealing to improve fabrication yield and sensor performance. The nanoscale infrared photodetector was characterized by digital microscopy and precise linear stage in order for fully understanding it. Besides, the low noise, high gain readout system was designed together with CNT photodetector to make the nano sensor IR camera available.To explore more of infrared light, we employ compressive sensing algorithm into light field sampling, 3-D camera and compressive video sensing. The redundant of whole light field, including angular images for light field, binocular images for 3-D camera and temporal information of video streams, are extracted and expressed in compressive approach. The following computational algorithms are applied to reconstruct images beyond 2D static information. The super resolution signal processing was then used to enhance and improve the image spatial resolution. The whole camera system brings a deeply detailed content for infrared spectrum sensing.
Show less
Title
Scalable pulsed mode computation architecture using integrate and fire structure based on margin propagation
Creator
Hindo, Thamira
Date
2014
Collection
Electronic Theses & Dissertations
Description
Neuromorphic computing architectures mimic the brain to implement efficient computations for sensory applications in a different way from that of the traditional Von Neumann architecture. The goal of neuromorphic computing systems is to implement sensory devices and systems that operate as efficiently as their biological equivalents. Neuromorphic computing consists of several potential components including parallel processing instead of synchronous processing, hybrid (pulse) computation...
Show moreNeuromorphic computing architectures mimic the brain to implement efficient computations for sensory applications in a different way from that of the traditional Von Neumann architecture. The goal of neuromorphic computing systems is to implement sensory devices and systems that operate as efficiently as their biological equivalents. Neuromorphic computing consists of several potential components including parallel processing instead of synchronous processing, hybrid (pulse) computation instead of digital computation, neuron models as a basic core of the processing instead of the arithmetic logic units, and analog VLSI design instead of digital VLSI design. In this work a new neuromorphic computing architecture is proposed and investigated for the implementation of algorithms based on using the pulsed mode with a neuron-based circuit.The proposed architecture goal is to implement approximate non-linear functions that are important components of signal processing algorithms. Some of the most important signal processing algorithms are those that mimic biological systems such as hearing, sight and touch. The designed architecture is pulse mode and it maps the functions into an algorithm called margin propagation. The designed structure is a special network of integrate-and-fire neuron-based circuits that implement the margin propagation algorithm using integration and threshold operations embedded in the transfer function of the neuron model. The integrate-and-fire neuron units in the network are connected together through excitatory and inhibitory paths to impose constraints on the network firing-rate. The advantages of the pulse-based, integrate-and-fire margin propagation (IFMP) algorithmic unit are to implement complex non-linear and dynamic programming functions in a scalable way; to implement functions using cascaded design in parallel or serial architecture; to implement the modules in low power and small size circuits of analog VLSI; and to achieve a wide dynamic range since the input parameters of IFMP module are mapped in the logarithmic domain.The newly proposed IFMP algorithmic unit is investigated both on a theoretically basis and an experimental performance basis. The IFMP algorithmic unit is implemented with a low power analog circuit. The circuit is simulated using computer aided design tools and it is fabricated in a 0.5 micron CMOS process. The hardware performance of the fabricated IFMP algorithmic architecture is also measured. The application of the IFMP algorithmic architecture is investigate for three signal processing algorithms including sequence recognition, trace recognition using hidden Markov model and binary classification using a support vector machine. Additionally, the IFMP architecture is investigated for the application of the winner-take-all algorithm, which is important for hearing, sight and touch sensor systems.
Show less
Title
Artificial lateral line systems for feedback control of underwater robots
Creator
AbdulSadda, Ahmad
Date
2012
Collection
Electronic Theses & Dissertations
Description
A lateral line system, consisting of arrays of flow-sensing neuromasts, allows fish and amphibians to probe their ambient environment and plays a vital role in their behaviors spanning predator/prey detection, schooling, rheotaxis, courtship and communication. The feats of biological lateral lines have inspired an increasing interest in engineering a similar sensing module for underwater robots and vehicles. Often known as artificial lateral lines, these sensors could potentially enable an...
Show moreA lateral line system, consisting of arrays of flow-sensing neuromasts, allows fish and amphibians to probe their ambient environment and plays a vital role in their behaviors spanning predator/prey detection, schooling, rheotaxis, courtship and communication. The feats of biological lateral lines have inspired an increasing interest in engineering a similar sensing module for underwater robots and vehicles. Often known as artificial lateral lines, these sensors could potentially enable an underwater robot to detect and identify moving or stationary objects, and exploit ambient flow energy for efficient locomotion. Despite the advances made in this area, realizing a practical artificial lateral line still faces significant challenges in both signal processing and flow sensor fabrication.In this dissertation we describe our effort in developing signal processing methods for hydrodynamic object localization and tracking using an artificial lateral line (ALL). We consider two types of objects, a vibrating sphere and a non-vibrating cylinder, both of which are of interest in underwater applications. A vibrating sphere, known as a dipole source, is widely used in the study of biological lateral lines and it emulates the rhythmic body or fin movement. A non-vibrating cylinder (with unknown cross-section shape), on the other hand, represents a general moving or stationary object underwater with a 2D flow profile. First, a novel bio-inspired artificial lateral line system is proposed for underwater robots and vehicles by exploiting the inherent sensing capability of ionic polymer-metal composites (IPMCs). Analogous to its biological counterpart, the IPMC-based lateral line processes the sensor signals through a neural network, and we demonstrate the performance of this approach in the localization of a dipole source. Second, with an assumption of potential flows, we formulate nonlinear estimation problems for the localization and tracking of a dipole source based on analytical flow models, and propose and compare several algorithms for solving the problem. For the case of a moving cylinder, we use conformal mapping to represent a general cross-section profile, and explore the use of Kalman-filtering-type techniques in the tracking and size/shape estimation of the object.We have conducted extensive experiments to validate the developed algorithms with an artificial lateral line prototype made of millimeter-scale IPMC sensors, with sensor-to-sensor separation of 2 cm, which is determined through an optimization process based on the Cramer-Rao bound (CRB) analysis. Finally, we experimentally explore the use of IPMC sensors for estimating the hydrodynamic parameters involved in a Karman vortex street that is created by a stationary cylinder in a flow. We validate that the vortex shedding frequency, which can be extracted from the sensor signal, shows clear correlation with the flow speed and the obstacle size.
Show less
Title
Data-driven and task-specific scoring functions for predicting ligand binding poses and affinity and for screening enrichment
Creator
Ashtawy, Hossam Mohamed Farg
Date
2017
Collection
Electronic Theses & Dissertations
Description
Molecular modeling has become an essential tool to assist in early stages of drug discovery and development. Molecular docking, scoring, and virtual screening are three such modeling tasks of particular importance in computer-aided drug discovery. They are used to computationally simulate the interaction between small drug-like molecules, known as ligands, and a target protein whose activity is to be altered. Scoring functions (SF) are typically employed to predict the binding conformation ...
Show moreMolecular modeling has become an essential tool to assist in early stages of drug discovery and development. Molecular docking, scoring, and virtual screening are three such modeling tasks of particular importance in computer-aided drug discovery. They are used to computationally simulate the interaction between small drug-like molecules, known as ligands, and a target protein whose activity is to be altered. Scoring functions (SF) are typically employed to predict the binding conformation (docking task), binary activity label (screening task), and binding affinity (scoring task) of ligands against a critical protein in the disease's pathway. In most molecular docking software packages available today, a generic binding affinity-based (BA-based) SF is invoked for the three tasks to solve three different, but related, prediction problems. The vast majority of these predictive models are knowledge-based, empirical, or force-field scoring functions. The fourth family of SFs that has gained popularity recently and showed potential of improved accuracy is based on machine-learning (ML) approaches. Despite intense efforts in developing conventional and current ML SFs, their limited predictive accuracies in these three tasks have been a major roadblock toward cost-effective drug discovery. Therefore, in this work we present (i) novel task- specific and multi-task SFs employing large ensembles of deep neural networks (NN) and other state-of-the-art ML algorithms in conjunction with (ii) data-driven multi-perspective descriptors (features) for accurate characterization of protein-ligand complexes (PLCs) extracted using our Descriptor Data Bank (DDB) platform.We assess the docking, screening, scoring, and ranking accuracies of the proposed task-specific SFs with DDB descriptors as well as several conventional approaches in the context of the 2007 and 2014 PDBbind benchmark that encompasses a diverse set of high-quality PLCs. Our approaches substantially outperform conventional SFs based on BA and single-perspective descriptors in all tests. In terms of scoring accuracy, we find that the ensemble NN SFs, BsN-Score and BgN-Score, have more than 34% better correlation (0.844 and 0.840 vs. 0.627) between predicted and measured BAs compared to that achieved by X-Score, a top performing conventional SF. We further find that ensemble NN models surpass SFs based on other state-of-the-art ML algorithms. Similar results have been obtained for the ranking task. Within clusters of PLCs with different ligands bound to the same target protein, we find that the best ensemble NN SF is able to rank the ligands correctly 64.6% of the time compared to 57.8% obtained by X-Score. A substantial improvement in the docking task has also been achieved by our proposed docking-specific SFs. We find that the docking NN SF, BsN-Dock, has a success rate of 95% in identifying poses that are within 2 Å RMSD from the native poses of 65 different protein families. This is in comparison to a success rate of only 82% achieved by the best conventional SF, ChemPLP, employed in the commercial docking software GOLD. As for the ability to distinguish active molecules from inactives, our screening-specific SFs showed excellent improvements over the conventional approaches. The proposed SF BsN-Screen achieved a screening enrichment factor of 33.90 as opposed to 19.54 obtained from the best conventional SF, GlideScore, employed in the docking software Glide. For all tasks, we observed that the proposed task-specific SFs benefit more than their conventional counterparts from increases in the number of descriptors and training PLCs. They also perform better on novel proteins that they were never trained on before. In addition to the three task-specific SFs, we propose a novel multi-task deep neural network (MT-Net) that is trained on data from three tasks to simultaneously predict binding poses, affinities, and activity labels. MT-Net is composed of shared hidden layers for the three tasks to learn common features, task-specific hidden layers for higher feature representation, and three outputs for the three tasks. We show that the performance of MT-Net is superior to conventional SFs and competitive with other ML approaches. Based on current results and potential improvements, we believe our proposed ideas will have a transformative impact on the accuracy and outcomes of molecular docking and virtual screening.
Show less
Title
Cortex-inspired developmental learning networks for stereo vision
Creator
Solgi, Mojtaba
Date
2013
Collection
Electronic Theses & Dissertations
Description
How does the human brain make sense of the 3D world while its visual input, the retinal images, are only two-dimensional? There are multiple depth-cues exploited by the brain to create a 3D model of the world. Despite the importance of this subject both for scientists and engineers, the underlying computational mechanisms of the stereo vision in the human brain is still largely unknown. This thesis is an attempt towards creating a developmental model of the stereo vision in the visual cortex....
Show moreHow does the human brain make sense of the 3D world while its visual input, the retinal images, are only two-dimensional? There are multiple depth-cues exploited by the brain to create a 3D model of the world. Despite the importance of this subject both for scientists and engineers, the underlying computational mechanisms of the stereo vision in the human brain is still largely unknown. This thesis is an attempt towards creating a developmental model of the stereo vision in the visual cortex. By developmental we mean that the features of each neuron are developed, instead of hand-crafted, so that the limited resource is optimally used. This approach helps us learn more about the biological stereo vision, and also yields results superior to those of traditional computer vision approaches, e.g., under weak textures. Developmental networks, such as Where-What Networks (WWN), have been shown promising for simultaneous attention and recognition, while handling variations in scale, location and type as well as inter-class variations. Moreover, in a simpler prior setting, they have shown sub-pixel accuracy in disparity detection in challenging natural images. However, the previous work for stereo vision was limited to 20 pixel stripes of shifted images and unable to scale to real world problems. This dissertation presents work on building neuromorphic developmental models for stereo vision, focusing on 1) dynamic synapse retraction and growth as a method of developing more efficient receptive fields 2) training for images that involve complex natural backgrounds 3) integration of depth perception with location and type information. In a setting of 5 object classes, 7 × 7 = 49 locations and 11 disparity levels, the network achieves above 95% recognition rate for object shapes, under one pixel disparity detection error, and under 10 pixel location error. These results are reported using challenging natural and synthetic textures both on background and foreground objects in disjoint testing.
Show less
Title
Capacity assurance in hostile networks
Creator
Li, Jian
Date
2015
Collection
Electronic Theses & Dissertations
Description
Linear network coding provides a new communication diagram to significantly increase the network capacity by allowing the relay nodes to encode the incoming messages. However, this communication diagram is fragile to communication errors and pollution attacks. How to combat errors while maintaining the network efficiency is a challenging research problem. In this dissertation, we study how to combat the attacks in both fixed network coding and random network coding.For fixed network coding,...
Show moreLinear network coding provides a new communication diagram to significantly increase the network capacity by allowing the relay nodes to encode the incoming messages. However, this communication diagram is fragile to communication errors and pollution attacks. How to combat errors while maintaining the network efficiency is a challenging research problem. In this dissertation, we study how to combat the attacks in both fixed network coding and random network coding.For fixed network coding, we provide a novel methodology to characterize linear network coding through error control coding. We propose to map each linear network coding to an error control coding. Under this mapping, these two codes are essentially identical in algebraic aspects. Meanwhile, we propose a novel methodology to characterize a linear network coding through a series of cascaded linear error control codes, and to develop network coding schemes that can combat node compromising attacks. For random network coding, we propose a new error-detection and error-correction (EDEC) scheme to detect and remove malicious attacks. The proposed EDEC scheme can maintain throughput unchanged when moderate network pollution exists with only a slight increase in computational overhead. Then we propose an improved LEDEC scheme by integrating the low-density parity check (LDPC) decoding. Our theoretical analysis, performance evaluation and simulation results using ns-2 simulator demonstrate that the LEDEC scheme can guarantee a high throughput even for heavily polluted network environment.Distributed storage is a natural application of network coding. It plays a crucial role in the current cloud computing framework in that it can provide a design trade-off between security management and storage. Regenerating code based approach attracted unique attention because it can achieve the minimum storage regeneration (MSR) point and minimum bandwidth regeneration (MBR) point for distributed storage. Since then, Reed-Solomon code based regenerating codes (RS-MSR code and RS-MBR code) were developed. They can also maintain the MDS (maximum distance separable) property in code reconstruction. However, in the hostile network where the storage nodes can be compromised and the packets can be tampered with, the storage capacity of the network can be significantly affected.In this dissertation, we propose a Hermitian code based minimum storage regenerating (H-MSR) code and a Hermitian code based minimum bandwidth regenerating (H-MBR) code. We first prove that they can achieve the theoretical MSR bound and MBR bound respectively. We then propose data regeneration and reconstruction algorithms for the H-MSR code and the H-MBR code in both error-free networks and hostile networks. Theoretical evaluation shows that our proposed schemes can detect the erroneous decodings and correct more errors in the hostile network than the RS-MSR/RS-MBR code with the same code rate respectively.Inspired by the novel construction of Hermitian code based regenerating codes, a natural question is how to construct optimal regenerating codes based on the layered structure like Hermitian code in distributed storage. Compared to the Hermitian based code, these codes have simpler structures and are easier to understand and implement. We propose two optimal constructions of MSR codes through rate-matching in hostile networks: 2-layer rate-matched MSR code and m-layer rate-matched MSR code. For the 2-layer code, we can achieve the optimal storage efficiency for given system requirements. Our comprehensive analysis shows that our code can detect and correct malicious nodes with higher storage efficiency compared to the RS-MSR code. Then we propose the m-layer code by extending the 2-layer code and achieve the optimal error correction efficiency by matching the code rate of each layer's MSR code. We also demonstrate that the optimized parameter can achieve the maximum storage capacity under the same constraint. Compared to the RS-MSR code, our code can achieve much higher error correction efficiency. The optimized m-layer code also has better error correction capability than the H-MSR code.
Show less
Title
Design and implementation of integrated self-powered sensors, circuits and systems
Creator
Huang, Chenling
Date
2011
Collection
Electronic Theses & Dissertations
Description
Wireless sensor systems have been widely used for both industrial and civil applications. With the development of circuit design and fabrication technique, sensor nodes now can be implemented with small scale at low cost, which is promising for ubiquitous sensing. However, with more functions integrated, the conflict between power consumption and expected lifetime became critical. Sensor nodes powered with batteries are generally compromised by extra physical size and periodic battery...
Show moreWireless sensor systems have been widely used for both industrial and civil applications. With the development of circuit design and fabrication technique, sensor nodes now can be implemented with small scale at low cost, which is promising for ubiquitous sensing. However, with more functions integrated, the conflict between power consumption and expected lifetime became critical. Sensor nodes powered with batteries are generally compromised by extra physical size and periodic battery replacement. Therefore, energy harvesting techniques are intensively involved in sensor design where environmental signal acts as auxiliary energy source.A typical energy harvesting sensor consists of four parts: energy harvester, energy storage, power management and sensor subsystem. Energy harvester scavenges power from environmental signal which is then transferred into energy storage. Since the output power is usually not in appropriate form, power management is used to provide a usable supply voltage/current for sensor subsystem. The limitation of energy harvesting sensor is determined by the power consumption of sensor subsystem, the efficiency of energy conversion and the available energy level from environment.In this dissertation, a novel solution referred as "self-powered sensor" is proposed to extend the limitation of energy harvesting sensor. The proposed sensor can directly harvest energy from input signal being sensed. Therefore the usage of energy storage and power management is eliminated, which achieves higher energy efficiency.To demonstrate proposed solution, the system and circuit design of a self-powered sensor are presented for long-term ambient vibration monitoring. Constrained by its application, the sensor can only scavenge energy from input strain signal itself, in which scenario all existing energy harvesting techniques fail. The greatest design challenge is to achieve both ultra-low power computation and non-volatile storage. In this dissertation, a novel technique based on floating-gate transistor is presented. By exploiting controllable hot electron injection procedure, specific computation can be performed according to the characteristic of input signal. In addition, floating-gates can also retain computation results with no power consumption.For autonomous sensing, a hybrid energy harvesting topology is proposed on system level. The sensor is designed with two different operation modes. In self-powered sensing mode, it can perform continuous monitoring, computation and data storage which is powered by input strain signal. In data interrogating mode, additional functions such as data sampling and wireless communication can be enabled once a certain reading device is provided.The dissertation is organized as follows. In chapter 1, the history of wireless sensor system is reviewed. The motivation of self-powered sensor and the contributions of this dissertation are presented. Existing energy harvesting techniques are evaluated in chapter 2. In chapter 3, the case of long-term ambient vibration monitoring is studied and the hybrid energy harvesting topology is proposed for self-powered sensor system. In chapter 4, the principle of ultra-low power computation and non-volatile storage is explored based on controllable injection procedure on floating-gate transistor. To verified proposed solution, a sensor prototype was fabricated in 0.5-um standard CMOS process. The details of circuit design and evaluation are presented in chapter 5, including analog signal processor, analog-to-digital converter, radio frequency front-end, digital baseband, etc. Chapter 6 shows an extension of ultrasonic powering and communication system based on preliminary work and chapter 7 draws final remarks.
Show less
Title
Automated Speaker Recognition in Non-ideal Audio Signals Using Deep Neural Networks
Creator
Chowdhury, Anurag
Date
2021
Collection
Electronic Theses & Dissertations
Description
Speaker recognition entails the use of the human voice as a biometric modality for recognizing individuals. While speaker recognition systems are gaining popularity in consumer applications, most of these systems are negatively affected by non-ideal audio conditions, such as audio degradations, multi-lingual speech, and varying duration audio. This thesis focuses on developing speaker recognition systems robust to non-ideal audio conditions.Firstly, a 1-Dimensional Convolutional Neural...
Show moreSpeaker recognition entails the use of the human voice as a biometric modality for recognizing individuals. While speaker recognition systems are gaining popularity in consumer applications, most of these systems are negatively affected by non-ideal audio conditions, such as audio degradations, multi-lingual speech, and varying duration audio. This thesis focuses on developing speaker recognition systems robust to non-ideal audio conditions.Firstly, a 1-Dimensional Convolutional Neural Network (1D-CNN) is developed to extract noise-robust speaker-dependent speech characteristics from the Mel Frequency Cepstral Coefficients (MFCC). Secondly, the 1D-CNN-based approach is extended to develop a triplet-learning-based feature-fusion framework, called 1D-Triplet-CNN, for improving speaker recognition performance by judiciously combining MFCC and Linear Predictive Coding (LPC) features. Our hypothesis rests on the observation that MFCC and LPC capture two distinct aspects of speech: speech perception and speech production. Thirdly, a time-domain filterbank called DeepVOX is learned from vast amounts of raw speech audio to replace commonly-used hand-crafted filterbanks, such as the Mel filterbank, in speech feature extractors. Finally, a vocal style encoding network called DeepTalk is developed to learn speaker-dependent behavioral voice characteristics to improve speaker recognition performance. The primary contribution of the thesis is the development of deep learning-based techniques to extract discriminative, noise-robust physical and behavioral voice characteristics from non-ideal speech audio. A large number of experiments conducted on the TIMIT, NTIMIT, SITW, NIST SRE (2008, 2010, and 2018), Fisher, VOXCeleb, and JukeBox datasets convey the efficacy of the proposed techniques and their importance in improving speaker recognition performance in non-ideal audio conditions.
Show less