應用於行動健康照護之智慧感測

Abstract

隨著物聯網（IoT）與穿戴裝置已衍然成為下一個新興市場的浪潮，提供多重感測模組收集大量生理資訊來用於相關的監控應用程序，並使判讀決策變得更佳可靠是非常必要的。然而多重感測器方法導致在感測平台和無線傳輸上，需求更高的數據傳輸率和更大的功率消耗。因此，本論文提出可應用在行動健康照護監控平台之一款低功率事件驅動讀出電路和多款新興的生理感測器，與具備特徵提取之感測器相互參照架構/系統的的解決方案與關鍵技術。 對於讀出電路與感測器上的改進，本論文著重在高解析度，低功耗，和高能量效率的應用要求上。再者，多款可應用於行動健康照護監控平台之新興的生理監控感測器將在論文中被提出，特別的是，一款可應用在生醫檢測上之極具潛力的實驗室晶片（FPLOC）將被開發與討論。最後，本論文展示一款具特徵提取之感測器相互參照架構/系統的行動健康照護監控平台，其中心電圖信號/呼吸狀態/胸腔起伏變化/當下的運動行為/與生醫檢測可以用於更準確的病徵判讀上數據交叉參考。這意味著本論文對於行動健康照護監控的應用上可降低更多的感測器功耗與無線傳輸功率和節省數據傳輸頻寬。 針對監測人體運動的行為，並實現低功耗與高能源效率之讀出電路，一款基於事件驅動之電容式三軸硬體共用，並具有因應補償微機電後製程電容誤差之自動修補電路將被發表。所提出的3軸的讀出電路使用聯電0.18微米製程實現其電路佔0.0354平方毫米面積，而補償微機電後製程電容誤差之自動修補電路面積為0.2204平方毫米。實驗結果中顯示在1.8V供應電壓下，單軸的電路功耗為50微瓦(FOM=1.92pJ)，若使用三軸硬體共用讀出架構，三軸同時在採樣頻率為125KHz讀出率下，僅消耗82微瓦(FOM=1.03pJ) ，並提供0.1g的加速度感測解析度，其最小可感測到±0.1fF微機電加速度計電容變化量。其償微機電後製程電容誤差之自動修補電路將可以0.06fF的修補解析度自動補償128級電容誤差，亦指可以校正±7.68fF 微機電電容失配問題。 對於呼吸監測，一款事件驅動型、節能型、以比例為基礎的濕度傳感器將被發表(當中，更多克服製程-電壓-溫度變異的技術與超快的濕度監控反應速度將被再次被改善)，所發表的濕度計使用台積電0.35微米製程，實現了15.6有效轉換位元，45.8微瓦平均功耗，20-90％相對濕度感測範圍，達成在1KS/s的採樣頻率，0.02％相對濕度靈敏度與超快10毫秒的濕度監控反應速度下，其每一位元讀出僅消耗6.1pJ能量，在溫度與電壓變異中，此款濕度感測器可將溫度變異所產生的40％相對濕度錯誤率下降至0.2％相對濕度錯誤率，亦可將電壓變異所產生的50％相對濕度錯誤率下降至0.1％相對濕度錯誤率。測量結果表明，該發表的濕度傳感器具有高靈敏度，超快響應時間和具競爭力的能源效率，這些特色使得它非常適合用於穿戴式呼吸監測上。 對於生物醫學檢測，一款智能的數位微流體處理器將發表呈現。這個實驗室晶片是從電路設計到架構規劃至系統自動化以及應用層面所發展完成，所述實驗室晶片實現了液滴操作、致動、液珠位置感測和電容式測量窗口的功能。再者，其可自我測試電路的正確性、不正常的操作電極位置、後製程後的晶片表面平整性與多種液珠的位置感測。這種新穎的實驗室晶片雛型解決許多傳統上的發展瓶頸並實現簡易操縱、易於監視、系統自動化和適用於生醫檢測所需的高精準度感測需求。使用標準的0.35微米CMOS製程製作，本論文中呈現兩款的30x30與30x60微電極之實驗室晶片雛型，並採用所發表的高解析度、低功耗之讀出電路來當成電容式反應觀測窗。測量結果顯示，微液珠可以如使用者規劃所操作，而且液珠位置感測的電容分辨率是1.3fF和可感測到0.39fF電容變化之高解析度讀出電路。這些實驗結果充分突顯出，本論文所提出之智能型數位微流體處理器可以以非常有效的方式被用於大量的生物檢測。 對於本論文所提出的行動健康照護監控應用中，本論文發展一款感測器相互參照架構與方法，以提供一個節能與數據可靠的解決方案。經採用事件驅動讀出電路架構與先前提及的感測器，其能量效率是可以再次提升;另一方面，由於多重生理感測器提供可相互參考的資料，其病徵判讀的正確性可以被提升。由其結果所示，其感測器功耗與無線傳輸功率和數據傳輸頻寬將可被降低並提供更好的判讀分析精確度，有關本論文所展示的行動健康照護監控測試平台亦證明了實施的可行性，在效能比較上，對於本行動健康照護監控平台將可達到超過40%的整體功耗改善。 由上述這些實施結果比較，意味著本論文所提出的關鍵技術，如讀出電路，新興感測器和感測器相互參照架構與方法，用於行動健康照護應用係非常有效率的。

As Internet of Things (IoT) and wearable devices become the next wave of emerging markets, it is very necessary to provide a set of sensors related to target applications so that decision-making becomes more reliable from collected data. However multi-sensor approach results in higher data rate and larger power consumption in sensing platform and wireless transmission. Hence, the low power event-driven readout circuit, emerging sensors, and sensor-fusion architecture/system solution with feature extraction are presented with the applied key techniques for mobile healthcare applications. The improvements of readout circuit and sensors are considered for high resolution, low power, and high energy efficiency requirements. Also, more emerging physiological sensors are explored for mobile healthcare applications in our proposal. Besides, a field programmable lab-on-a-chip (FPLOC) is developed for biomedical detection. Finally, the proposed mobile healthcare platform with sensor-fusion solution is demonstrated, where ECG signal/ respiration status/ chest cavity downs/ corresponding human motions/ biomedical detection can be employed for data cross-reference. This implies our proposal for mobile healthcare applications can save more transmission power and data bandwidth in wireless module. To monitor human motion behavior and achieve low power and high energy efficiency of readout circuit, 3 axis-shared capacitive readout circuit based on event-driven architecture with capacitance auto-trimming circuit for the MEMS accelerometer is proposed. The proposed 3-axis readout circuit with 0.0354mm2 area is fabricated in UMC 0.18μm CMOS-MEMS process, and trimming circuit of the MEMS capacitance compensating occupies 0.2204mm2. Experimental results show power consumption is 50μW with 1.8V supply voltage for 1-axis (FOM=1.92pJ), 82μW for 3-axis (FOM=1.03pJ) under 125KHz of sampling frequency and 0.1g acceleration sensitivity for ±0.1fF MEMS capacitance change. The auto-trimming can compensate ±7.68fF MEMS capacitance mismatch via the 0.06fF resolution with 128 adjustable stages. For respiratory monitoring, an event-driven, energy-efficient, proportion-based humidity sensor is proposed (also, overcoming PVT variation technology and ultra-fast response speed are considered) . This work achieves 15.6b 45.8μW 20-90 %RH at 1KS/s, 6.1pJ per sample, 0.02%RH of sensitivity, and 10ms fast response time in TSMC 0.35μm CMOS-MEMS process. With variations in temperature and voltage, our proposal can minimize the errors from 40%RH to 0.2%RH and 50%RH to 0.1%RH. Measurement results show the proposed humidity sensor has high sensitivity, ultrafast response time and competitive energy-efficiency, making it very suitable for wearable respiratory monitoring. For biomedical detection, an intelligent digital micro-fluidic processor is presented. From circuit, architecture, system, to application, the work integrates the functions of droplet control, actuation, location sensing and capacitive measurement window. Further BIST for circuitry, faulty microelectrodes, chip flatness after post-fabrication, and droplet category classification are achieved. This novel prototype solves lots of traditional development bottlenecks to implement the easy-to-control, easy-to-monitor, system automation and high accuracy sensing for bioassay detection purposes. Fabricated in standard 0.35μm CMOS process, two prototypes occupy 30x30 and 30x60 microelectrodes with measurement window, where the high resolution capacitive readout circuit is employed. Measured results show droplet examines can be functioned and the sensitivity of location detection is 1.3fF and 0.39fF for high resolution readout circuit. These experimental results highlight that our proposal can be used for the huge amount of bioassays and big-data analysis in a very efficient and effective way. For the proposed mobile healthcare application, a sensor-fusion approach is introduced to provide an energy-efficient and data-reliable solution. By exploiting event-driven architecture and mentioned sensors, energy efficiency can be enhanced; on the other hand, analysis accuracy can be further improved with the support of multi-data sets. As a result, both power consumption and data bandwidth can be minimized with better accuracy to meet those specifications in battery-powered devices. The test vehicles related to mobile healthcare applications will also be introduced to demonstrate the feasibility of our proposal. With evaluation, our proposal can enhance over 40% improvement on total power consumption in the proposed mobile healthcare platform. These preliminary results demonstrate that our proposed key techniques, such as readout circuit, emerging sensors, and sensor-fusion architecture, can be applied to mobile healthcare applications very effectively and efficiently.