• 學位論文

基於信譽模型之點對點視訊傳輸獎勵機制

點對點視訊傳輸系統在現今的網際網路中已相當普及，使用者自願性的提供頻寬以利系統架構出點對點傳輸系統。然而，自私使用者只下載卻不上傳任何檔案的行為導致其餘使用者的視訊品質不穩定甚至讓點對點系統崩解。面對這類型的問題，現存的研究方法利用過去在點對點檔案分享系統上成功的獎勵機制，移轉至點對點視訊串流傳輸系統架構上，這些方法的特色是採用周圍鄰近使用者對我的貢獻程度，作為我分配資源的依據條件。但是並非所有類型的點對點視訊傳輸系統都適用於這種主觀依據自己得到的好處來分配資源的獎勵機制。推拉式的點對點視訊傳輸架構中，使用者互相提供頻寬資源的狀況並非常見的傳輸狀態，現有的獎勵方法利用相互間的貢獻度分配資源在推拉式的點對點傳輸架構下造成使用者壁壘分明無法互相傳輸視訊封包的窘況。我們利用信譽評等的方式在推拉式的點對點視訊傳輸架構下建立獎勵機制，使用者替自己以外的鄰近使用者評分，累積這些分數得到每個使用者對點對點視訊傳輸系統的整體貢獻度，系統內每個使用者都可以得到這項資訊做為是否同意其他使用者不同視訊層級傳輸要求的依據，根據我們的機制可以讓不同貢獻度的使用者有不同的視訊品質並壓抑自私使用者的視訊品質至相當低的水平。

• 學位論文

Intelligent Dynamic Camouflage System

偽裝可以分為兩種類別，分別是自然偽裝與人工偽裝。枯葉蝶是自然偽裝的一個例子，其顏色與紋理和形狀如同牠常出沒的枯木枯葉，使其不易被他的敵人發現，達到生存機率上升的目的。人工偽裝的概念源自於這種動物自我保護的自然偽裝機制。好的人工偽裝不僅僅可以使軍人存活下來，也是打贏一場戰役的重要因素。 傳統的人工偽裝方式多以在待偽裝物上以人工的方式繪製或覆蓋上與週遭環境類似的顏色或圖案，造成敵方觀察者在視覺上的錯覺，以達到欺瞞偽裝的目的。然而傳統的偽裝方式在待偽裝物移動至不同背景區域，通常其效不彰。因此，我們建立一套智能化的動態偽裝系統，使待偽裝物可以隨著地點變化自然融入週遭環境中。 在這篇論文中，我們使用一台UBOT（機器人移動平台）提供動態環境。在其上架設兩台相機分別用來取得前方關於觀察者的資訊和被遮蔽的背景資訊，一台筆記型電腦用來模擬待偽裝物，利用螢幕來顯示背景圖樣。我們提出一個動態偽裝的系統，此系統包含了一個估計觀察者深度與方位的子系統，以及一個找尋並顯示合適背景圖樣的子系統。 我們的系統可以有效的隨著待偽裝物的移動顯示出合適的圖片，達到動態偽裝的目的。雖然目前還無法做到即時偽裝，但花的時間仍在可接受的範圍。

• 學位論文

針對動態重組單晶片網路的新穎架構與路由演算法

Hybrid networks, which combine the flexibility of packet-switched network and the low-latency of circuit-switched network, provide decent performance. However, existing researches require users to maintain the configuration of hybrid networks. Users need to spend a lot of time and effort on analyzing the applications at first. Then, they can know how and when to configure the hybrid network. The exhausting process motivates us to construct a hybrid network which reconfigures itself automatically. In this thesis, we propose Dynamic Bypass Circuit, a novel architecture which reconfigures the network according to the traffic trends of applications at run-time. Therefore, users need not to worry about the mapping between the applications and the network. Additionally, in order to make the trends more obvious, we modified original north-last routing algorithm into north-last-weave routing algorithm. It guides packets toward different directions according to the position of their source nodes. The simulation results show that dynamic reconfigurable NoC which adopts Dynamic Bypass Circuit outperforms the traditional packet-switched network. The latency improvement is 18.08% under Transpose traffic and on average 10.76% under singular synthetic traffic patterns.

• 學位論文

利用生物網路標記進行肺癌分子研究及診斷

Lung cancer is the leading cause of cancer deaths worldwide. Many studies have investigated the carcinogenic process and identified the biomarkers for signature classification. However, to the best of knowledge, there is no satisfactory network-based method for carcinogenesis characterization and diagnosis from the system perspective. In this study, a systems biology approach, which integrated microarray gene expression profiles and protein-protein interaction information, was proposed to develop a network-based biomarker for molecular investigation and diagnosis of lung cancer. The network-based biomarker consists of two protein association networks constructed for cancer samples and for non-cancer samples. Based on the network-based biomarker, a total of 40 significant proteins are identified with carcinogenesis relevance values (CRVs) to gain an insight into the network mechanism of lung carcinogenesis. In addition, the network-based biomarker is also acted as the diagnostic tool, demonstrated to be effective to diagnose the smokers with lung cancer. The results show that the network-based biomarker through constructed protein association networks successfully sheds light on the pathways and mechanisms in lung carcinogenic process and, most importantly, provides potential therapeutic targets to combat against cancer.

• 學位論文

懸浮閘電晶體於0.18μm單層多晶矽CMOS製程之開發與其在類比電路上的應用

在積體電路發展的歷程中，懸浮閘元件經常被用來製作非揮發性( Nonvolatile )記憶體，如EPROM、快閃記憶體等。由於懸浮閘元件可相容於標準CMOS製程中，九○年代開始，有許多研究則將其做為設計類比電路時的一種方法。近年來，隨著製程技術的日益精進，懸浮閘元件的製作方式也必須有所改變，本論文提出一個可於0.18μm單層多晶矽CMOS製程中製作的懸浮閘元件，並且能夠將其應用於類比電路設計上。 本論文提出的懸浮閘元件結構，係由三個PMOS所組成，元件的寫入操作採用離子化熱電子注入與帶對帶穿隧電流，抹除操作則為FN 穿隧效應，同時亦嘗試以通道熱電洞注入機制做為抹除操作。本論文驗證元件在寫入與抹除操作時的電性變化，元件耐久度與資料保久度的實驗數據也會在論文中提出。 為了能夠以SPICE模擬懸浮閘元件在電路操作時的特性，本論文透過熱電子注入電流與FN穿隧電流模型的建立，提出此元件的等效電路模型；並將實驗測量結果與模擬結果相互比較，驗證此模型的近似效果。本論文進一步將懸浮閘元件用於類比記憶體電路的設計，期望能透過電路的操作，適性地控制懸浮閘電壓，並將此元件推展至其他電路的應用。 最後，針對Translinear電路在標準CMOS製程中製作的諸多限制，本論文應用懸浮閘元件提出一個新的Translinear 電路設計方案，藉由改變電晶體的電壓電流轉換關係，期望能夠減小Translinear電路在操作時的誤差；並以電流公式的推導與電路模擬結果，驗證此方案的可行性。

• 學位論文

Capture Power Reduction for At-Speed Scan-Based Testing Using Layout-Aware Location X-Filling

With the rapid improvement of testing technique, the basic of testing function is a necessary technique. It also needs to reduce the testing time and power dissipation as well as testing cost. Nowadays, it is important to solve problem which is the more power dissipation during testing. The more power dissipation can cause the effect of IR-drop and hot-spots. The chip is easy to fail with the high power dissipation. Therefore, applying the concept of gate density, this thesis proposes two methods (Probable Weighted Switching Activity X-Fill, PWSA X-Fill; and Location-Aware X-Filling) to reduce the power dissipation and distributes the accumulated heat. Applied in ISCAS-89 benchmark circuits (s9234, s13207, and s38417), the simulation results show that the original power is reduced about 10% peak and 20% average in the launch cycle and 9% peak and 12% average by the proposed x-fill methods. Compared with other x-filling, the proposed x-filling has better result on reducing launch and shift cycle power and would reduce the occurrence of IR-drop and hot-spots for at-speed scan-based testing.

• 學位論文

非線性隨機分佈參數系統的強健控制設計：使用模糊方法

In this dissertation, we address the robust stabilization design problem for nonlinear stochastic distributed parameter systems (NSDPSs) with random external disturbances and measurement noises in the spatio-temporal domain. A fuzzy stochastic distributed parameter system is proposed to approximate the NSDPS based on fuzzy interpolation approach. Then a fuzzy stochastic spatial state space model is developed to represent the fuzzy stochastic distributed parameter system via a semi-discretization finite difference scheme. Based on this model, a robust fuzzy estimator-based stabilization controller is proposed to stabilize the NSDPS. Furthermore, the robust stochastic $H_infty$ stabilization design is proposed to attenuate the effects of random external disturbances and measurement noise in the spatio-temporal domain from the area energy point of view, and the LMI technique is applied to solve the control gains and estimator gains of the controller via a systematic control design procedure. Finally, a simulation example is given to illustrate the design procedure and to confirm the performance of the proposed robust fuzzy estimator-based stabilization design for the NSDPSs.

• 學位論文

Modeling and Implementation of a Module Integtrated Converter for Photovoltaic Systems

有鑑於全球受到能源安全、氣候變遷、與永續發展的挑戰，近年來有效的使用再生能源已成為全球矚目的焦點。在眾多的再生能源中，太陽能光伏系統在環境上與經濟效益上擁有較多的優勢相較於其他系統。為了有效利用太陽能資源,太陽能交流模組系統近期最被廣泛的研究與探討。所謂模組整合型轉換器即為能夠提供隨插即用的功能的直交流轉換器，如此太陽光能才能被最有效率的使用。 由於太陽能光伏模組具有低輸出電壓的特性，所以本論文提出一應用於建築整合型太陽光電系統高升壓型擬直流鏈模組整合轉換器。基本上，本論文主要貢獻總結如下所述。首先，針對建築整合型太陽光電系統提出一單級具電器隔離擬直流鏈模組整合轉換器以解決遮蔽效應的問題。藉由耦合電感特性以及低頻切換直交流轉換以提升本論文模組整合轉換器整體效率。除此之外，採用本實驗試所研發之太陽能最大功率追蹤技術達到瞬時最大功率控制。第二，藉由轉換器的直流與交流數學模型推導作為穩態分析與閉迴路補償的依據。最後，本論文實際製作一輸入電壓30V、輸出電壓110Vrms、額定功率150W的雛型系統。整體的轉換效率都高於88%以上。由模擬與實驗的結果驗證本論文所提轉換器的有效性。

• 學位論文

Robust Sensorimotor Control of Human Arm System under State-dependent Noises, Control-dependent Noises and Additive Noises

The robust control of human arm movements is planed according to the integration of sensory information, sensory motor transformation and human brain computation. The human arm is controlled to an acceptable posture in a robust optimal way. Due to the intelligent nature of human judgments, the application of Takagi-Sugeno (T-S) fuzzy model to human judgments is appealing to emulate the intelligent computation in the prefrontal cortex of human brain in the sensorimotor control of human arm movement. Here, we try to apply a robust fuzzy estimator-based control scheme to mimic the sensorimotor control of realistic planar movement of the human arm. However, the state variables of the planar model of the human arm are not all available via visual and proprioception information. With the help of posture (state) estimation based on human sensory information in human brain, a robust fuzzy estimator-based control is introduced for sensorimotor reference tracking control of arm movement in spite of internal noises, state-dependent noises, environmental noises and uncertain initial values. Based on fuzzy interpolation for nonlinear stochastic arm system, the complicated noise-tolerance robust control of human arm tracking problem can be simplified by solving a set of linear matrix inequalities (LMIs) through Newton’s iterative method via an interior point scheme of convex optimization. Finally, a simulation example is given to illustrate the control procedure and confirm the performance of the robust fuzzy estimator-based sensorimotor control for human arm system.

• 學位論文

低通影像處理對人眼視覺中隨機共振現象之影響

關於人眼視覺方面的文獻探討在最近已有越來越多的趨勢，而連帶的也越來越多影像處理的討論以此為研究基礎。藉由知覺心理學方面的實驗，學者們確信人眼視覺中的效應和視覺影像的品質是息息相關的，因此探討人眼視覺中關於雜訊影像所引發的效應成為了一個重要的議題。 近年來一個針對雜訊方面的訊號處理的方式漸漸受到注意，那便是隨機共振。這個訊號處理演算法的概念在於，加了雜訊的訊號反而比沒加上雜訊的訊號更為容易被辨識出原訊號。而這樣的現象在生理訊號處理以及影像相關領域中已有不少相關的探討。此外人眼視覺的隨機共振效應也經由相關的知覺心理學實驗得到證實，並且衍生相關的研究。 對於影像去雜訊以及壓縮相關技術而言，低通影像處理占了非常重要的地位。甚至一些影像品質評估的演算法也是基於此開發而成的。加了雜訊後的影像經過低通影像處理後所產生的動態影像，經過人眼視覺測試後其隨機共振的效應會有甚麼樣的變化是本論文主要探討課題，文中將會藉由使用和客觀的靜態影像品質分析的對照做一個分析討論。