RESEARCH
優秀學生作品
優秀專題研究
111年上學期獲獎研究專題
Two-Stage Amplifier Design for Bio-signal Acquisition(指導教授 廖育德 教授)
The method for designing a two-stage amplifier involves applying the circuit design techniques acquired from the Analog Integrated Circuit and Electronics courses. Through this process, we analyze the various circuit behaviors, focusing on how the length and width of the MOS affect its performance. It was observed that the different designs result in varying bias voltages and operating regions, among other factors. Thus, precise calculation of parameters is necessary to determine the appropriate MOS sizing for the amplifier to achieve optimal performance. We opted to design a two-stage amplifier because a one-stage amplifier would entail a tradeoff between gain and swing, whereas a two-stage configuration can address both requirements separately.
Two-stage OP-AMP Design and Simulation (指導教授 陳科宏 教授)
We have designed a two-stage operational amplifier using Hspice in order to meet the necessary specifications in the last week. Since the OPAMP is a crucial and fundamental circuit within the realm of analog integrated circuits, and It is also utilized in a range of applications, including biomedical, ADC, power PMIC, and more, we determined that the OPAMP design would be an appropriate topic for our poster.
 
資料傳輸基於光通訊設計之特性 (指導教授 陳巍仁 教授)
The optical communication system consists of a transmitter, a channel, and a receiver. The transmitter encodes information into an optical signal, and optical communication has many advantages, including: high speed transmission, high security (difficult to be interfered with), large distance transmission, and durability (optical fibers can withstand high tension and high temperature). Optical communication is widely used in broadband networks, telecommunication networks and other high-speed data transmission applications.
 
物件偵測模型於內視鏡影像息肉辨識實作 (指導教授 黃俊達 教授)
Since the colorectal cancer is one of the top three cancers with the highest mortality rates according
to WHO, polyp detection in colonoscopy images is undoubtedly an extremely important issue. Our task
involves refining the latest state-of-the-art object detector, YOLOv7, with a private de-identified dataset
provided by Chang Gung Memorial Hospital, which includes more than 2400 colonoscopy images. After
integrating several extra optimization techniques into the training flow and polishing the model
architecture slightly, we have successfully achieved significant improvement in terms of both precision and
recall as compared to the baseline model.

 
AFE Design with Folded Cascode Amplifier for Biomedical Signal Detection (指導教授 洪崇智 教授)
Design of CMOS Folded Cascode Amplifier for Biomedical Signal Detection (指導教授 洪崇智 教授)
Design of Logarithmic Amplifier and its Application on ESD Event Detection(指導教授 柯明道 教授)
Electrostatic discharge (ESD) is a phenomenon that can cause serious damage to
integrated circuits (ICs). Thus, an ESD-event detector has been designed and fabricated in a single chip to detect and alarm the ESD events in integrated circuit (IC) manufacturing environments. Therefore, it motivates us to improve the ESD-event detector, by identifying ESD events more quickly and accurately.

 
110年上學期獲獎研究專題
Dual-Source Energy-Harvesting Interface With Cycle-by-Cycle Source Tracking and Adaptive Peak-Inductor-Current Control (指導老師: 陳柏宏教授)
無人機障礙物影像辨識 (指導教授:李佳翰老師)
透過此專題的研究,針對無人機飛行競速訓練出最合適的即時影像辨識模型,能夠快速辨識出障礙物的位置,並且準確的標示出外框以避免無人機撞上邊框。此外,也提供不同影像辨識模型在準確度以及辨識速度等指標的分析比較,對整體效果以及實務上的應用進行綜合分析。
機器人在防疫旅館的應用-回收垃圾 (指導教授:楊谷洋老師)
使用者可利用Line通知機器人回收指定房間的垃圾,機器人在接收到指令時,如果處於待機狀態就會直接前往指定房間。到達目的地後,透過手臂上的RealSense amera找尋垃圾袋並計算座標,之後透過手臂夾取垃圾袋。
將其運送到存放地點放置。機器人如在工作期間接收到使用者的要求,會通知使用者稍待片刻,待目前工作執行結束後,再接著處理閒置的要求。機器人在沒有接收到使用者傳送指令的情況下,會回到待機位置,等待下一個指令。
基於通道狀態之室內人員定位模擬 (指導教授:方凱田老師)
利用CSI訊號模擬室內人員是否存在並且判斷其位置,並利用CNN及Dense兩種網路使得準確度最大化,以利於之後應用於實際Wi-Fi訊號上能判斷實際人員之定位。
將機器學習演算法部署於邊緣裝置並進行系統優化 吃進多少卡路里?AI技術告訴你! (指導教授:黃俊達老師)
隨著人們外食的次數日漸增加,常常將各式各樣的食物吃進肚子。由於餐廳不一定會提供餐點所含的熱量,所以人們就會不小心攝取太多,造成身體上的負擔。因此我們在這項專題中,運用機器學習(ML)中的影像辨識(image classification)分辨出食物的種類,估計其體積、重量和熱量,方便人們控管飲食。
Photoplethysmography System (指導教授:廖育德老師)
光體積變化描記圖訊號常被用於血氧血壓的偵測,因此設計電路系統來產生更明確的光體積變化描記圖訊號是非常重要的,本專題著眼於相關電路的改進和波型分析以提升相關效能。
110年下學期獲獎研究專題
Breath detection based ontransfer learning (指導老師: 方凱田教授)
無線區域網路實體層基頻收發端之系統 (指導教授:方凱田老師)
以802.11a作為傳輸介面時,MAC層往下傳遞給實體層時有固定格式,稱為 PSDU,接著再由實體層加上可供接收端判讀的 header 後傳送出去,一個 frame 則包含了 PSDU、header 等在一次傳輸中所有符元的集合。
當接收端收到一個 frame 後,依照 format 解碼、解調變、去除 header後即可獲得所需的資料 (payload),而frame format 可分成三大部分 :preamble、signal field、data field在此次專題,我們著重在解出接收到的訊號data,並藉由 signal fieldRATE (前4 bits) 得出調變方式與 code rate。
Radar Missing Dimension Recontruction from Motion (指導教授:王傑智老師)
針對移動偵測重建出雷達相對容易忽視的穩態物體,減少因為雷達系統錯誤產生的意外事故。
How to Fine-tune the Self-supervised Model? (指導教授:帥宏翰老師)
自監督機器學習技術用於室內定位的fine-tune領域建立模型,fine-tune分析出室內定位不同狀態的資訊變化,以產出一個可以藉由室內資訊變化得知物體實際位置的系統。
Implementation of Pedestrian Trajecotry Prediction Module-科技部自駕車次系統關鍵技術團隊 (指導教授:鄭文皇老師)
在無人駕駛系統中,軌跡浴廁扮演相當重要的一個環節,乃至於ADAS(先進駕駛輔助系統)上也有其應用的價值。若系統能提前理解周遭其他人車可能的運動趨勢,就能更充分的時間預防碰撞與重新規畫路徑決策,也有意味著能營造著更安全舒適的使用體驗。