关键词： Driving pattern recognition (DPR)   fuel cell (FC)/supercapacitor (SC) hybrid electric vehicle (HEV)   fuzzy energy management   genetic algorithm (GA)   neural network (NN) classifier  

摘要： A novel adaptive energy management strategy is proposed for real-time power split between fuel cells (FCs) and supercapacitors (SCs) in a hybrid electric vehicle in view of the fact that driving patterns greatly affect fuel economy. The driving pattern recognition (DPR) is achieved based on the features extracted from the historical velocity window with a multilayer perceptron neural network. After the DPR has been obtained, an adaptive fuzzy energy management controller is utilized for power split according to the required power for vehicle running. In order to prolong the FC lifetime while decreasing the hydrogen consumption, a genetic algorithm is applied to optimize critical factors such as adaptive gains and fuzzy membership function parameters for several standard driving cycles. In the proposed method, the future driving cycles are not required and the current driving pattern can be successfully recognized, demonstrating that less current fluctuations and fuel consumption can be achieved under various driving conditions. Compared with conventional energy management systems, the proposed framework can ensure the state of charge of SCs within the desired limit.

关键词： Location  

摘要： A systematic review across U.S. Geological Survey (USGS) identified potentially induced seismic locations was conducted to discover seismic distance patterns and trends over time away from injection disposal wells. Previous research indicates a 10 km (6 miles) average where the majority of induced seismicity is expected to occur within individual locations, with some areas reporting a larger radius of 35 km (22 miles) to over 70 km (43 miles). This research analyzed earthquake occurrences within nine USGS locations where specified wells were identified as contributors to induced seismicity to determine distance patterns from disposal wells or outward seismic migration over time using established principles of hydrogeology. Results indicate a radius of 31.6 km (20 miles) where 90% of felt earthquakes occur among locations, with the closest proximal felt seismic events, on average, occurring 3 km (1.9 miles) away from injection disposal wells. The results of this research found distance trends across multiple locations of potentially induced seismicity.

关键词： Landing safety   Flight data   Pattern recognition   Markov chain  

摘要： This paper aims to predict the landing state during the landing phase to ensure landing safety and reduce the accidents loss. Some past researches have demonstrated the landing phase is the most dangerous phase in flight cycle and fatal accident. The landing safety problem has become a hot research problem in safety field. This study concentrates more on the prediction and advanced warning for landing safety. Firstly, four landing states are divided by three flight parameter variables including touchdown, vertical acceleration and distance to go;subsequently, pattern recognition based on BP neural network is used to establish the landing state prediction model;the genetic algorithm is used to initialize the model parameter;the Markov chain is proposed to revise and improve the model for higher prediction precision. Finally, in comparison with pattern recognition and the Markov chain revision results, the Markov chain revision method is demonstrated to be practical and effective.

关键词： Singular value decomposition (SVD)   Canonical correlation analysis (CCA)   Motor unit action potential (MUAP)   Collective correlation coefficient matrix (CCCM)   Electromyography (EMG)  

摘要： A support system with efficient learning framework helps eliciting complete knowledge of underlying phenomena of interest. It makes the analysis less-onerous, time-consuming and error-prone and thus promotes large scale applications. Such modeling requires profound understanding of available information and its appropriate utilization. Albeit success of electromyogram (EMG) support systems, challenges still exits specifically in early phase of design mainly due to inherent variations and complex data distribution patterns of signals. In this article, a frame singular value decomposition (F-SVD) based method-generalizing Canonical correlation analysis for automatic classification of EMG signals to diagnose amyotrophic lateral sclerosis (ALS), myopathy and normal subjects, is proposed. At first, signals are decomposed to formulate a set of vectors and performed subspace transformation to demonstrate the variability and stability of signals base on correlations between pairs of vectors. Besides, discrete Wavelet transformation is applied on generated vectors and correlation analysis is performed. Afterwards, taking highly correlated statistical measures a set of compact feature distributions are estimated and fused via two recently proposed parallel and serial feature fusion models. Finally two global descriptors for effective classifications of various EMG patterns are proposed. The efficacy of derived feature space is validated by intuitive, graphical and statistical analysis. The model performances are investigated over two datasets. It achieves accuracy of 98.10% and 97.60% over two and three-class groups of first dataset receptively. Accuracy over second dataset is 100% with a specificity of 100% and sensitivities of 100%. This is first time that F-SVD is employed for automatic classification of EMG. Experiments results on various datasets evince adequacy of our method. Further comparison of performance with state-of-the-art methods depicts that our method compara

关键词： 分布式光纤定位系统   双Mach-Zehnder干涉仪   模式识别   时延估计算法  

摘要： 长期以来,周界安防问题已成为社会研究的热点,对非法入侵行为的及时发现和有效定位具有重要研究价值及实际意义。传统的安防系统极易受到外界环境的影响,而基于光纤传感技术的连续分布式光纤入侵定位系统具有灵敏度高、抗干扰、无死角、对扰动信号进行定位等优点。由于传感系统受噪声和外界环境影响,导致误报率高,定位精度低等问题。为了在复杂的环境中,能够高效的识别入侵行为,并进行快速准确定位,本文围绕分布式光纤入侵定位系统对振动信号的识别与定位算法展开深入研究。采用两级振动模式识别方法,并利用二次相关算法对噪声环境下入侵信号进行定位,实验结果表明,该算法减少了系统误报率,提高了定位精度。本论文的主要工作包括:(1)在对双Mach-Zehnder干涉型光纤传感器原理深入阐述的基础上,对振动信号的特征提取与识别算法进行了研究。将采集到的信号进行预处理后,通过对信号的短时能量与短时平均过零率设定阈值,提取出有效组,完成时域一级预判;然后分别对提取到的有效组信号以峭度和时频熵作为时域与时频域特征形成特征向量,并采用BP神经网络算法对攀爬、敲击和下雨信号进行多类模式识别,完成二级识别。通过对现场采集到的监测数据进行大量实验,证明了算法的有效性及可靠性。(2)针对光纤感知信号中噪声对定位精度影响,利用互相关算法、二次相关算法和广义二次相关算法对攀爬信号在不同的信噪比环境进行定位精度比较,实验结果表明,广义二次相关算法在低信噪比环境下,定位精度最高,能实现对人为破坏行为的准确定位,为外界复杂环境下安防系统定位的准确性与实时性提供了算法上的支持。(3)基于LabVIEW和MATLAB软件平台,利用模块化的设计思想实现了短时能量与短时过零率、经验模态分解与Hilbert变换、BP神经网络、广义二次相关以及定位等模块的程序编写,并完成了分布式光纤传感定位系统的设计,通过实验验证,证明了系统的可行性。

关键词： 人机交互   信息融合   模式识别   意图理解   柔顺控制  

摘要： 随着智能机器人技术的不断发展,人机交互已成为一个新的研究领域。赋予机器人强大的交互能力已成为提升人机交互表现的重要前提,尤其是在接触式人机协作中,赋予机器人强大的意图理解能力已成为提升接触式人机协作柔顺性的必由之路。然而,在接触式人机协作中,由于当前柔性皮肤触觉传感器的研究并不成熟,再加上神经电信号（脑电信号、肌电信号）信号微弱、抗干扰性差等缺点的制约,接触式人机协作中的意图理解已成为制约接触式人机交互中机器人柔顺性的关键问题。因此,针对当前接触式人机交互在意图理解和柔顺控制方面所存在的共性问题,本文以人与机器人共同完成轨迹跟随的协作任为背景,进行了基于触力觉融合信息的人机交互与柔顺控制方法研究,研究内容主要包括:（1）接触式人-人协作意图表达方式分析与感知系统设计以人-人协作为出发点,研究人-人协作机理,分析接触式人-人协作意图表达方式,从多方式交互的角度去解决接触式人机协作中的意图理解问题;同时完成了触觉和力觉感知系统的设计,并对触觉传感器进行了性能测试与封装。（2）基于触力觉融合信息的操作意图描述定义与决策生成为了对接触式人机协作中操作者的操作意图进行全面的描述与定义,在采用传统六维力传感信息的基础上,增加了实验室自主研发的电容阵列柔性触觉压力传感器,通过采集人在操作过程中手部抓握图像,根据传感器的压力分布和大小来间接体现人的操作意图;首先,采用多模态信息融合技术来获取、描述与定义操作者的操作意图;然后,采用卷积神经网络对人的操作意图模式进行识别;最后,根据描述与定义的操作意图及操作意图模式识别结果进行决策生成。（3）基于触力觉融合信息的接触式人机协作柔顺控制方法研究在上述操作意图描述定义与决策生成的基础上,提出了一种基于触力觉融合信息的接触式人机协作柔顺控制策略,将能够反映操作意图模式切换的变量引入到控制系统中,设计了一种基于触觉反馈的变导纳控制器,实时调整机器人的虚拟阻尼参数,以响应操作者的控制意图;搭建了基于触觉反馈的变导纳控制实验平台系统,进行了接触式人机协作随动实验。实验结果表明,基于触觉反馈的变导纳控制策略能够较好的“理解”人的操作意图,极大地提升了接触式人机协作的柔顺性。

关键词： 恶臭   特征提取   模式识别   定性分析   定量分析  

摘要： 随着我国经济的快速发展,环境污染问题愈发严重。恶臭污染作为一种特殊的环境污染,已成为世界七大公害之一,严重影响生态环境和人类身体健康。目前常用的恶臭检测方法有嗅觉测试法和仪器分析法,嗅觉测试法根据嗅辨人员的嗅觉对恶臭气体进行检测,具有主观性强、检测速度慢、实时性差等缺点;仪器分析法需要专用的设备进行检测,成本昂贵、实时性差、不能实现复合恶臭的化学组分、含量和恶臭气味的同步测试。因此,研发一套能够快速准确检测复合恶臭气体的系统,实现在线自动测量,具有重要的现实意义。通过对国内外现有恶臭检测方法的研究,并结合实际需要,研发了一套以电子鼻为核心的复合恶臭在线检测系统,并完成了传感器阵列组建、恶臭信号采集、气路设计、模式识别算法的设计,成功实现了复合恶臭气体的定量检测。恶臭定量检测结果受多种参数的影响,其中有效信号、特征值提取和模式识别算法是关键的三部分。有效的特征可以大大提高恶臭检测系统的性能,降低模式识别难度。本文采用分段拟合响应曲线的方式进行特征提取,并结合主成分分析法(PCA)和线性判别分析法(LDA)进行特征选择。模式识别算法是对提取后的特征进行适当处理,从而获得准确的恶臭浓度信息。合理的模式识别算法可以提高模式识别精度,本文对比研究了偏最小二乘法(PLS)和BP神经网络两种模式识别算法。实验结果表明,BP神经网络对恶臭气体的浓度预测有更高的精度,PLS对低浓度的恶臭气体无法建立良好的回归预测模型。为了评价恶臭对人类环境的影响,需要建立恶臭评价指标和恶臭浓度的响应模型。影响恶臭评价指标的因素不仅跟恶臭浓度相关,而且也跟恶臭成分相关。因此,在定量分析前必须先定性分析得到恶臭成分。对比研究了支持向量机(SVM)和BP神经网络两种定性识别方法,结果表明,LDA与SVM/BP神经网络的结合能100%识别恶臭气体。

关键词： 生物特征识别   指静脉   图像增强   特征提取   模式识别  

摘要： 随着信息科技的蓬勃发展以及人们生活水平的不断提高,对个人信息的保护和鉴定也越来越受到广泛的重视。在当今的网络信息化时代,需要鉴定个人身份的应用场景越来越多,并且要求对其进行数字化和隐私化处理。像指纹识别、虹膜识别这类传统生物特征识别技术仅能利用生物体的“表象”特征来进行个人身份的鉴定,安全系数较低。因此出现了一种更为先进、且独一无二的生物特征识别技术——指静脉识别技术。指静脉识别技术不仅改善了传统生物特征识别技术在实用性和普适性的痛点,而且在安全性上也得到了极大的提高。正是因为指静脉识别技术独特的优越性使其成为了该领域的研究热点。本研究中完成的工作如下:通过分析多张指静脉图像得出其纹路特点,提出了基于自适应多阈值的指静脉特征提取算法,并改进了传统的邻域方向模板,用以检测图像中指静脉横截面的谷形深度和曲率。然后将整幅指静脉图像划分成多块大小相等、互不相交的子图像,并根据给定的阈值将这些子图像分为指静脉特征区域、背景区域以及噪声区域。最后依据不同的灰度均值自适应地为每一块子图像设定的阈值,最大限度地从图像中提取出指静脉的特征信息。通过多组实验结果表明,使用本文算法提取出的指静脉纹路特征相较传统算法更加清晰且光滑,尽可能还原了指静脉纹路特征的实际拓扑结构,并能够更好地契合图像质量偏低的使用场景。作为指静脉识别技术核心算法的研发平台,指静脉识别仿真系统需要满足进行算法研究所必须的功能需求。第一部分从整个识别仿真系统最基本的两大功能模块注册和识别出发,为其设计了多个相应的子功能模块,包括指静脉图像的预处理模块、指静脉特征提取模块和指静脉特征匹配模块。第二部分是在.Net Framework框架的基础上配合OpenCV库函数完成了应用程序的开发,该程序不仅实现了指静脉识别仿真系统的基本功能,而且能够提供后台管理以及数据存储功能。

关键词： Translation studies   Computer engineering   Information technology   High energy physics  

摘要： Modern high energy particle physics experiments are characterized by large scale capture, storage, and analysis of data. A large amount of data is generated by these experiments. However, the capacity for data storage is limited, so smart data reduction by distinguishing interesting particle tracks from background data in real time is required. This is challenging because of extremely tight latency constraints and enormous amounts of background data. In this thesis, we present our work on the development and optimization of custom content addressable memory (CAM) based pattern recognition associative memory systems for fitting track detector data captured in high energy particle physics experiments. We present the complete design, simulation, and testing of a novel multi-Vdd CAM chip along with an exploration of the multi-Vdd design space. Following this, we propose the use of 3D IC technology to vertically integrate these cells to reduce parasitics and increase pattern density. Two 3D prototypes have been designed; one vertically integrates the same building blocks as the first 2D prototype, and the other uses an additional pipelined binary priority encoder design for output sparsification. Associated studies on the modeling of power consumption behavior of CAMs and the integration of thin-film thermocouples in 3D ICs for temperature measurement are also discussed. We conclude with the testing results of the first prototype of the precision timing ASIC for LGAD-based CMS endcap timing layer upgrade chip which investigates adding timing information along with location data for track fitting in future experiments.

关键词： Physical therapy   Biomedical engineering   Robotics  

摘要： This thesis presents work toward a novel rehabilitation tool for infants with limited arm movement such as those who sustain a perinatal brachial plexus injury (PBPI). PBPI is a traction injury to peripheral nerves that occurs during the birth process. An injury to the upper trunk of the plexus (C5-6 spinal nerves) partially or fully denervates the skin of the upper arm and muscles of the elbow (i.e., biceps, brachialis) and shoulder (i.e., Deltoid, Sternal Pectoralis Major, Rotator Cuff). Initially, PBPI is typically treated with Passive Range of Motion (PROM) and positioning led by a physical therapist or occupational therapist. If recovery is limited, nerve microsurgery is indicated by 6 months of age. Recovery in infants with PBPI varies from 38 to 80% of infants depending on the initial condition and severity, as well as the rate of reinnervation. Yet, through technology, there may be methods available to increase the rate of recovery, leading to greater use of the affected arm. Infants as young as 3 months of age have been found to increase arm and leg movements through a paradigm of contingent reinforcement (i.e. rewarding desired movement patterns with audiovisual stimulation such as an overhead mobile). Before a device can be fully constructed to provide contingent reinforcement for desired arm movements, those movements must be consistently detectable. Thus, for this thesis project, I studied automatic detection of the arm movements desired for young infants with PBPI using arm acceleration data. Acceleration data was acquired from a wrist-worn sensor as an adult volunteer moved her arm in the desired motion. I implemented a template-matching algorithm based on taking the dot-product of a moving window of 3D acceleration vectors with template acceleration patterns, where the templates were movement samples deemed targeted and rehabilitative by an experienced physical therapist. I found that the algorithm detected rehabilitative movements with an accuracy o