Chung Cheng University Institutional Repository
Not a member yet
889 research outputs found
Sort by
不同視角 IPCam 攝影機之視訊畫面同步技術 ;Different Point of View IPCam Frame Alignment
No full text[[abstract]]隨著IPCam的廣泛使用，傳統監視攝影機已逐漸被IPCam取代，越來越多的IPCam應用在監控上。然而，在多台 IPCam 以較大視角對相同背景進行取像時，很少文獻探討讓多台IPCam視訊的畫面同步問題。傳統攝影機係以外同步訊號來達成同步目的，本論文則探討以IPCam影像內容分析比對來做為畫面同步的方式。本論文研究利用每個畫面擷取的特徵點 (包括前景人物頭部點、手部點、背景點)，再配合子序列對子序列的軌跡比對方式，計算出對應點對間的轉換矩陣 (Fundamental matrix)及轉換誤差來判斷畫面間是否比對成功，並藉以估測兩個IPCam 視訊序列間的時間差異 (即畫面同步位置)。根據實驗模擬結果顯示，當同時採用頭部點及手部點為特徵點時，加入少量的背景點，能以較少的視窗長度即可以達到滿意的畫面同步成功率
高選擇性寬頻帶拒濾波器與吸收式寬頻帶拒濾波器;Design of Wideband Bandstop Filter with High-selectivity and Wideband Absorptive Bandstop Filter
No full text[[abstract]]本論文主要研究分為兩部份，高選擇性寬頻帶拒濾波器與具寬頻特性之吸收式帶拒濾波器。第二章為濾波器分析理論，將簡單介紹濾波器雛型與設計方法。接著第三章為高選擇性寬頻帶拒濾波器的研究，利用兩種帶拒濾波器特性，分別為使用四分之一波長耦合線且具有L型諧振腔之架構與傳統四分之一波長開路段產生傳輸零點架構做結合，利用高品質因子元件，能有效結合前者的高選擇性效果和後者的寬頻特色，所得到需求的響應特性。更進一步討論此架構不同階數與不同零點位置的設計，所帶來的效果，由ADS軟體進行模擬取得特性結果，且改善頻帶外的通帶特性，經由電磁模擬，在L型諧振腔採用扇形傳輸線做為最終架構，此項微調其目的為降低電路敏感性，章節末段提供設計實例，在衰減量40 dB時頻寬仍能達到14.75 %的頻寬百分比，實做電路尺寸為0.37λg*0.59λg，其中 λg為3 GHz的引導波長(guided wavelength)。第二部分為寬頻吸收式帶拒濾波器，從傳統反射式帶拒濾波器(reflective BSF)出發且架構以四分之一波長耦合線諧振腔設計，中心頻率為3 GHz與吸收頻帶為20 dB的電路雛形，利用集總式元件(lump element)產生有損諧振腔，且探討吸收器的特性，與探討除了增強偶合強度外可使頻寬拓寬的可能性，章節末段提供設計實例，在吸收頻帶可達21.98 %的頻寬百分比且頻帶內RL>19.98 dB，實做電路尺寸為0.19λg*0.56λg。
For modern wireless communication applications, the researches of filter play an important role since it can have an effect on the sensitivity of communications systems. In this thesis, two parts are presented. The first part is the wideband bandstop filter with high-selectivity at chapter 3. To achieve wideband and high-selectivity performance, a structure of the planar bandstop filter (BSF) with four sections which can control transmission zeros (TZs) to adjust the bandwidth of requiring stop band is presented. We employ two methods to produce the TZs, respectively coupling lines with L-resonators and through lines with open-circuit stubs. The S-parameter decay deeply and quick when the insertion loss next to stop band. And the presented type can replace the conventional type with a quarter wavelength open-ended transmission line resonator. by combine with L-shaped resonators.The second part is the absorptive bandstop filter with wideband performance at chapter 4. The initial design comes from 4-th order reflective BSF, and adds lump elements (resistance) to produce absorptive effect. Use formulas to calculate the impedance match for absorptive part. For improving narrow bandwidth performance, I utilize π-shaped resonators. Finally, two of researches get good performance with simulations
應用於高效率視訊編碼的頻寬偵測技術;Available Bandwidth Estimation Technology for High Efficiency Video Coding
No full text[[abstract]]隨著時代科技的進步，近年來影像傳輸技術也不斷進步，網際網路已經成為生活不可缺少的一部分，隨著智慧手機與平板電腦等行動裝置的普遍性，使得愈來愈多的相關應用程式移植到這些3C產品上面，也不斷推陳出新，網路多媒體的發展也越來越迅速，近年來像是網路直撥、影音廣告、網路通訊，這些裝置出現在生活中，頻寬偵測對多媒體傳輸的影響性就顯得非常重要了，可以讓用戶端達到高品質的服務。 本論文提出自動頻寬偵測，可以封裝影像串流的檔案，做為探測封包傳送，故在實作上不需再傳送大量額外探測封包，以免造成網路負擔。頻寬偵測中提出的頻寬計算方法，可提升頻寬偵測的準確性，最後實測得到，在有線網路環境下的頻寬準確度約為97.3%，無線網路環境下的頻寬準確度約為95.6%，相較於其他頻寬偵測工具準確。
In recent years, the service of multimedia transmission becomes universal due to the rapid development of wireless networks. Smart phone and tablet PC become more and more popularization, bring more convenience to people. The transmission of video streaming service for wireless is becoming increasingly popular, and most have a compressed video stream into different specifications to meet users for different bandwidth demand. However, bandwidth detection mechanism is very important in the transmission of video streaming. Using the information of bandwidth detection will make the clients to be high-quality services.The proposed bandwidth detection mechanism, which method is that encapsulate the video stream files become probing packet, and it do not need any additional burden. It has advantage of both active and passive measurement of detection. In this paper, the method used to calculate the bandwidth of self-made, so you can upgrade the client bandwidth detection accuracy. For the easily packet loss of wireless network environment, the methods we adopted, avoid the large amount of error from bandwidth detection due to packet loss, it can improve the accuracy. Our methods are proved to be superior to other bandwidth detection tools in measured results
具自我激磁加速系統穩定之微瓦級風力能量採集系統;A Self-Excitation Wind Energy Harvesting System with Settling Enhancement Using Micro Turbine
No full text[[abstract]]本篇論文完成一應用於風力輸入之能量採集系統並具有能量擷取提升及快速啟動電路。環境中的風力可轉換成給予電路工作的能量，能量經過穩壓後給予外部電路做應用，但小型的換能器所能採集的能量大小有限，在不穩定的風源中除了一般風速的考量外，微弱風下的考量也是必要的。本論文使用微瓦能量等級之風力發電機(以下簡稱風機)作為能量來源，風機內部的寄生電感透過本論文的電源管理系統達到自激磁( Self-excitation )的功能。透過風機自激磁的方式解決交流輸入下低輸入電壓時的壓降限制( IR-drop )，提升輸入能量已達到快速充電啟動與能量提升的功能，本系統於低風速( < 14 Hz )環境下，自我激磁啟動電路之啟動速度高於二極體與蕭特基二極體橋式整流電路。此外，傳統的能量採集系統為了達到交流轉直流、最大能量採集與穩壓的功能，會分成三級功率轉換，增加了導通與開關損耗。本系統僅使用兩級功率轉換來達到功率消耗的減少。本系統在輸入尖端電壓2 V、輸入能量2 mW時擁有63 %的端點對端點( End to end )效率。最後，本論文使用台灣積體電路公司0.35 μm 2P4M CMOS製程，以100 CQFP封裝，尺寸為2.4×3.2 mm2。
A novel wind energy harvesting (WEH) system using micro-watt turbine is proposed in this paper. Toward enhancing the wind velocity range and system settling time of the WEH system, parasitic inductance of the turbine is utilized for self-excitation. The power converter in this system consists of an efficient two-stage converter rather than typical three-stage architecture. A regulator and the maximum power extraction (MPE) are also implemented in the same chip. The smallest input voltage to settle the system is less than 0.25 V and the settling time is less than traditional full-bridge rectifier with a 1 mF rectified capacitor. This work is implemented in TSMC 0.35-μm process, achieving 63 % end-to-end efficiency. The total active area is 7.68 mm2
利用類神經模糊邏輯輔助卡爾曼濾波器之即時定位與地圖構建;Simultaneous Localization and Mapping withNeuro-Fuzzy Assisted Extended Kalman Filtering
No full text應用於第五代行動通訊之領結陣列天線與電磁平面鏡設計;Design of Bow-tie Phased Array Antennas and Electromagnetic Surfaces for 5th-Generation Mobile System
No full text[[abstract]]本論文設計一應用於第五代行動通訊之相位陣列天線，參考華為(Huawei)提出的5G系統原型將陣列天線之操作頻率設計於3.5 GHz，天線單元採領結天線，並將其擴展為由64個天線單元組成之4×16相位陣列天線，以增加波束指向性與提升陣列增益。利用掃描單元場型量測法(Scanning Element Pattern Measurement Approach)經由量測陣列中各天線單元場型進而合成出陣列波形，其量測結果分別為：陣列最大增益為20 dBi，此外，當相位差改變範圍從?160度至+160度時，波束掃描範圍可從?57度至+57度。論文另一主題為印刷電路板電磁平面鏡設計，採用擁有寬可調反射相位範圍之具延遲線微帶電磁單元來增進電磁平面鏡的設計彈性，利用電磁模擬軟體驗證可降低交叉極化的子陣列中電磁單元擺放模式。基於印刷電路板反射面陣列天線的設計理論獲取每一單元所需之反射相位，藉由改變電磁單元之延遲線長度實現對應之反射相位，而達到波束成形之效果，並通過電磁模擬軟體驗證實際整體電磁平面鏡之特性。最後為檢測所研製電磁平面鏡可用性，針對選定之入射角度(30度與45度)進行不同反射角度之電磁平面鏡實現與量測。量測與電磁模擬結果呈現相當一致的吻合度。
In this thesis, the phased array antennas applicable to the fifth generation (5G) mobile systems are designed. Based on the 5G radio prototype implemented by Huawei, the operation frequency of the proposed phased array antennas is selected at sub-6GHz (i.e. 3.5 GHz). The bow-tie antenna is utilized as the antenna element constituting the 4×16 phased array antenna to enhance the beam directivity and increase the array gain required by 5G communication.Through measuring the gain pattern of individual element composing the array, the beam-forming and beam-steering characteristic can then be synthesized by scanning element pattern measurement approach. The measurement results show the maximum array gain is 20 dBi. In addition, when the progressive phase shift between the adjacent antenna is adjusted from ?160° to +160°, the corresponding beam directions can be scanned from +57° to ?57° thus validating the beam-steering. The other topic of this thesis is the design and implementation of printed-circuit board electromagnetic surfaces. The microstrip electromagnetic elements with delay lines possessing wide adjustable reflection phase range are utilized to improve the design flexibility of the electromagnetic surfaces. Based on the theory of the reflectarray, the phase distribution of elements can be acquired and the required reflection phase shift can be fulfilled by adjusting the physical delay lines thereby achieving the beam-forming capability. The characteristics of the electromagnetic surfaces are verified by full-wave electromagnetic simulator. Finally, to validate the practicality of the designed electromagnetic surfaces, several surfaces with different reflection angles are designed under two predetermined incident angles (30° and 60°). The measurement and simulation results show satisfactory agreement
相異使用者的評價於推薦系統之研究;A Study of Leveraging Dissimilar User Information in a Recommendation System
No full text[[abstract]]推薦系統主要在推薦使用者可能感興趣的資訊，如何在眾多資訊中找到最準確的推薦方法，並且有效率地給出推薦為一大課題。在推薦的過程中往往會遇到兩種問題，一是新使用者剛加入這個系統，系統要如何去推薦給新使用者資訊的冷起始問題；另一個則是使用者所評價過的物品數過少，導致系統對使用者的偏好不夠清楚，進而無法準確推薦的稀疏問題。在解決冷起始問題上，以往作法是把資料放入演算法進行訓練前，將相似使用者的評價也一併加入訓練，本研究則是探討在以往未納入考量的相異使用者，他們的評價是否也能對系統預測的準確率產生貢獻。將前處理好的資料放入多種演算法，最後以推薦系統的準確率作為驗證。本研究使用MovieLens電影評價資料庫，運用使用者已有的評價紀錄及新增的資料，藉由LIBMF演算法及Librec系統的SVD++、Slopeone演算法做推薦預測。
Recommendation systems mainly recommend a user with information which the user may be interested in. How to find the most accurate information and also an efficient way to recommend the user is a major issue. In the process of recommendation, there are often two problems existing in collaborative filtering. The lack of information of a new user or a new item may cause a problem in finding similar users or similar items. This is called cold start problem. The sparsity problem arises when not enough ratings are provided to the items to recommend. This study uses the MovieLens film evaluation database, and our method uses the user's existing rating records to make a prediction. In order to solve the cold start problem, a common practice is to put the similar users’ information into collaborative filtering. Our proposed method is to use the dissimilar users’ information to provide additional information that may contribute to the accuracy of the system. By examining the pre-processed data into a variety of algorithms, we compare the results of our approach
多重代理人模擬於虛擬都市中車輛與行人相互影響之研究;A Multiagent Approach to Vehicle-Pedestrian Interaction in the Virtual City
No full text[[abstract]]本研究為透過多重代理人模擬發展出以虛擬都市為核心的車輛與行人互動的系統，跟大部分的交通模擬不同，使用者除了當作觀察者的角色外，也可以成為交通系統中的用路人之一，與其它代理人互動。系統代理人種類有車輛代理人、行人代理人、交通號誌代理人、車輛產生代理人、行人產生代理人，其中車輛代理人擁有保守、一般與魯莽個性，行人代理人也擁有不同的行為模式。模擬系統使用遊戲引擎開發，擁有良好的畫面呈現、觀察機制，透過觀察面板使用者可以迅速得知需要的資訊，攝影機分為自由角度、車輛角度與行人角度視角。模擬環境中的道路分為直行道路、行人橋直行道路、90度轉彎、T字路口、十字路口，使用者可以觀察到車輛代理人與行人代理人、交通號誌代理人在路口之間細微的互動關係。
This thesis proposed a multiagent simulation approach to simulate the interaction between pedestrians and vehicles in a virtual city. Unlike most traffic simulations, the user who is the observer may also become a participant in the simulation world, and the user is able to interact with other agents. In the simulation system, there are vehicle agents, pedestrian agents, traffic light agents, vehicle generator agents and pedestrian generator agents. The vehicle agent has the personality type such as conservative, general and reckless type, and the pedestrian agent has different behavior patterns. In order to have better graphics and observation mechanism, the system is developed with the game engine. The user can quickly get the information they need through the observation panel, and the camera in the system is designed for three observing types, which are free view, car-follow and pedestrian-follow. The types of roads in the system are divided into straight road, straight road with pedestrian bridges, 90-degree turn road, T-junction, and intersection. With the system, user can observe the subtle interaction with vehicle agents, pedestrian agents and traffic light agents in the intersections
基於深度學習之3D視訊品質評估技術;No-reference 3D video quality assessment Based on convolutional neural network
No full text[[abstract]]近年來，隨著 3D 電視、電影的興起，3D技術受到學術界和業界的廣泛關注，越來越多人認同未來將是3D立體視訊技術快速成長的階段。 然而，由於構建3D觀看環境的困難，以及解決諸如視覺不適等各種3D質量因素，3D-VQA (3D Video Quality Assessment)的研究相對較少，因此，評估3D視訊品質的要求是迫切需要的。本論文係針對彩色加深度格式的3D視訊品質進行評估。現今深度估測演算法眾多，深度估測好壞對3D立體影像品質的影響大，例如當深度邊緣與彩色影像的邊緣無法對齊，或是深度圖估測錯誤時，會造成3D立體影像在觀看時前景物凹向畫面內或背景物凸出畫面的不良效果，所以要評斷所產生深度圖造成的3D立體影像主觀品質好壞為本論文主要目的。本論文針對數種深度估測方法，首先利用DIBR技術獲得左右眼立體影像，接著利用主觀評分方式取得立體影像的主觀分數，不同深度估測方法獲得不同主觀分數。接著利用四種視覺特徵做為輸入，例如:運動向量圖、彩色及深度邊緣圖…等，接著利用深度學習中的卷積神經網絡架構來做主觀評估分數的訓練及預測。深度學習部分採用卷積神經網絡是因為卷積神經網絡是一種由一個或多個卷積層和全連通層組成的學習架構，對於圖像處理或分類有出色表現。本論文整體架構為無參考式的客觀品質評估，意欲使深度學習架構所預測的3D視訊品質能接近人眼主觀品質，藉以判斷特定已訓練深度估測方法所產生深度圖的好壞或未經訓練深度估測方法的優劣。經實驗結果發現:本論文所提出的架構在實驗一，也就是評估已訓練深度估測方法所產生深度圖的好壞上，準確率可以達到70%以上。在實驗二，未經訓練深度估測方法的評估，其準確度可以達到80%~90%。不過也發現了一些問題，主觀分數樣本太過集中而導致樣本少的分數區間判斷錯誤。
In recent years, with the rise of 3D TV, the film, 3D technology by the academic community and the industry's attention, more and more people agree that the future will be 3D stereo video technology rapid growth stage. However, due to the difficulty of building a 3D viewing environment and solving various 3D quality factors such as visual discomfort, 3D-VQA’s (3D Video Quality Assessment) research is less, so the assessment of 3D video quality requirements is urgently needed.This paper evaluates the 3D video quality of the color plus depth format. The depth of the current estimation algorithm, the depth of the assessment of the quality of 3D stereo image quality, such as when the depth edge and color edge can not be aligned, or the depth map estimate error, it will cause 3D stereoscopic images When the foreground material concave to the screen or the background of the protrusion of the adverse effects of the screen, so to assess the depth of the 3D subjective quality score is our main purpose of this paper.In this paper, several kinds of depth estimation methods are used, we obtain the left and right eye stereoscopic images by using DIBR technique. Then, the subjective scores of stereo images are obtained by subjective scoring method. Different subjective scores are obtained by different depth estimation methods. And then use the four visual feature as input, such as: motion vector, color and depth edge map ... and so on, and then use the deep learning about convolution neural network architecture to do the subjective assessment of the scores of training and prediction. The deep learning part uses the convolution neural network because the convolution neural network is a learning architecture composed of one or more convolution layers and a fully connected layer, which is excellent for image processing or classification. The overall structure of this paper is a non-reference objective quality assessment, which means that the 3D video quality predicted by the deep learning architecture can be close to the subjective quality of the human eye in order to judge whether the depth map of the trained depth estimation method is good or bad Advantages and disadvantages of training depth estimation method.The experimental results show that the proposed scheme is more than 70% on the experimental one, that is, the depth map of the trained depth estimation method. In experiment 2, the accuracy of the untrained depth estimation method can be as high as 80% to 90%. But also found some problems, subjective scores of the sample too concentrated and lead to less sample interval judgment error
功率放大器輸出端主動變壓器結構設計與氮化鎵功率放大器線性度分析;Active Transformer Matching Design of Power Amplifier and Linearity Analysis of GaN Power Amplifier
No full text[[abstract]]本論文包括兩部分電路設計，第一部分是以CMOS 90 nm製程設計一應用在50 GHz (V Band)之功率放大器，輸出端利用具結合功率效果之變壓器結構阻抗轉換器設計。第二部分以氮化鎵高電子移動率電晶體設計一具效率與線性度之功率放大器，此功率放大器使用由AB類與C類放大器所組成之Doherty架構。第一部分設計的CMOS毫米波功率放大器其輸出匹配利用變壓器之特性設計功率結合器與阻抗轉換器，其原因為在高阻抗轉換比的情況下，傳統LC匹配網路之損耗會隨之增加，而變壓器之損耗與阻抗轉換比為獨立的，而本論文設計之變壓器可將兩組差動對放大器之功率結合並輸出，模擬電路之損耗在50GHz時為1.7 dB。而此電路之量測結果在50 GHz時電路增益為7.8 dB，輸出1 dB增益壓縮點為5.7 dBm，而飽和輸出功率為10.2 dBm。小訊號參數S21在50 GHz到67 GHz之間皆大於8 dB。第二部分設計的電路為Doherty功率放大器，由於傳統功率放大器操作在高效率區時線性度很差，反之操作在高線性區時效率也不會好，而此架構則是利用輸出端之主動式負載修正技術，以及結合AB類與C類放大器之特性，使得功率放大器輸出飽和點回退6 dB時還能保持高效率的特性，且不失線性度。本論文中設計出一具效率與線性度之功率放大器，其結果為當雙頻量測時輸出功率為36 dBm時的三階交互調變失真為-30 dBc以下，單頻輸出功率37 dBm時效率為41.9 %。
This thesis includes two parts, the first part is a 50 GHz (V band) power amplifier designed by CMOS 90 nm process. The output matching circuit is a transformer based impedance converter with a power combining structure. The second part is focused on the design of a GaN HEMT-based power amplifier with high efficiency and high linearity. The architecture is Doherty power amplifier, which is consisting of Class AB and Class C amplifiers.In the first part, the output matching of the CMOS millimeter wave power amplifier using the characteristics of the transformer plays as a power combiner and impedance converter. As in the case of high impedance-conversion ratio, the loss of the traditional LC matching network will be higher, and the loss of transformer is independent on impedance conversion ratio. The transformer not only converts the impedance but also combine the deleivered power of two sets of differential amplifiers, and the simulated loss of the transformer is about 1.7 dB at 50 GHz. This power amplifier as measured at 50 GHz, the circuit gain is 7.8 dB, the output 1 dB gain compression point is 5.7 dBm, and the saturated output power of 10.2 dBm. The s-parameter measurement shows the gain being greater than 8 dB between 50 GHz and 67 GHz.The second part of this thesis is the design of Doherty power amplifier. As the traditional power amplifier operation in the high efficiency region when the linearity is poor, otherwise, operating in the high linearity region, the efficiency of power amplifier will not be good. Doherty architecture uses active load modification at output port, and combines the characteristics of Class AB and Class C amplifier. So that the overall characteristic of power amplifier can maintain high efficiency without losing the linearity under the output power back off 6 dB from the saturation point. The measured result indicates that the third-order intermodulation distortion is lower than -30 dBc when the output power of the two-tone measurement is 36 dBm, and the one-tone output power is 37 dBm when the efficiency is 41.9 %