本論文使用機器學習技術研製一款穿戴式智慧紅外線腕式血壓感測器。該感測器包含紅外線感測光源、藍芽模組、訊號處理及機器學習演算法，並應用於收集31位受測者，其年齡介於20至30歲間，共計收集47筆有效手腕之橈動脈脈波訊號。再之，將原始脈波訊號使用MATLAB進行濾波、脈波獨立切分與特徵萃取。其中，特徵萃取步驟係將各獨立脈波使用傅立葉級數展開，取其正弦與餘弦係數作為後續機器學習之訓練特徵；將該特徵正規化後，則可輸入一維卷積類神經網路進行訓練。相較於市售醫療級血壓計，可獲得收縮壓之平均差(mean difference, MD)與標準差(standard deviation, SD)結果為2.36 ± 7.88 mmHg、舒張壓結果為1.31 ± 6.43 mmHg。

A wearable smart infrared wrist blood pressure sensor is developed in this thesis. This sensor is composed of infrared light source, Bluetooth module, signal processing, and machine learning algorithm. It is applied to collect 47 sets of effective wrist pulses upon the radial arterial area of 31 subjects, which are aged from 20 to 30. Moreover, the measured pulses are then filtered, pulse segment, and feature extraction by using MATLAB. Here, the features used for machine learning are extracted from coefficients of Fourier series expansion of each wrist pulse, namely the coefficients of sine and cosine basis functions. After normalizing the extracted features, they are sent into the one-dimensional convolution neural network (CNN) to perform model training. As compared with commercial sphygmomanometer, the mean difference (MD)±standard deviation (SD) of the calculated systolic blood pressure (SBP) is 2.36 ± 7.88 mmHg and MD ± SD diastolic blood pressure (DBP) is 1.31 ± 6.43 mmHg.

[1] 衛生福利部國民健康署, 高血壓是中老年人的專利，年輕人不需要量血壓?, 2019,
[Online]. Available: https://www.hpa.gov.tw/Pages/Detail.aspx?
nodeid=127&pid=6951
[2] 衛生福利部國民健康署, 慢性病盛行率, 2020,
[Online]. Available: https://www.hpa.gov.tw/Pages/Detail.aspx?
nodeid=641&pid=1231
[3] 衛生福利部統計處, 108年死因統計結果分析, 2020,
[Online]. Available: https://dep.mohw.gov.tw/dos/cp-4927-54466-113.html
[4] 衛生福利部國民健康署, 「您血壓量了沒？」定期量血壓全家健康有保障！, 2020,
[Online]. Available: https://www.hpa.gov.tw/Pages/Detail.aspx?
nodeid=3804&pid=10708
[5] 臺大醫院醫學工程部, 生理監視器功能與參數量測原理簡介, [Online]. Available:
https://www.ntuh.gov.tw/BMED/Fpage.action?muid=3899&fid=3681
[6] OMRON, OMRON血壓計HEM-7310 HEM-1000, [Online]. Available:
https://www.omronhealthcare.com.tw/product/list.php?index_prm_id=1
[7] ASUS VivoWatc h BP(HC-A05), [Online]. Available:
https://www.asus.com/tw/Mobile/Wearable-Healthcare/VivoWatch/ASUS-VivoWatch-
SP-HC-A05/
[8] T.-Y. Tu, and P. C.-P. Chao, “Continuous blood pressure measurement based on
a neural network scheme applied with a cuffless sensor,” Microsystem
Technologies, vol. 24, no. 11, pp. 4539-4549, Nov, 2018.
[9] R. K. Pandey, T.-Y. Lin and P. C.-P. Chao, “Design and implementation of a
photoplethysmography acquisition system with an optimized artificial neural
network for accurate blood pressure measurement,” Microsystem Technologies,
2021.
[10] C.-H. Tseng, T.-J. Tseng, and C.-Z. Wu, “Cuffless blood pressure
measurement using a microwave near-field self-injection-locked wrist pulse
sensor,” IEEE Trans. Microw. Theory Techn., vol. 68, no. 11, pp. 4865 -
4874, Nov, 2020.
[11] C.-Y. Chen, W.-K. Wang, T. KAO, B. C. Yu, and B. C. Chiang, “Spectral
analysis of radial pulse in patients with acute, uncomplicated myocardial
infarction,” Japanese Heart Journal, 1993, vol. 34, no. 2, pp. 131-143.
[12] Wang N., Yu Y., Huang D., Xu B., Liu J., Li T., Xue L., Shan Z., Chen Y.,
and Wang J., “Pulse diagnosis signals analysis of fatty liver disease and
cirrhosis patients by using machine learning,” The Scientific World Journal,
vol. 2015, Article ID 859192, 28 Nov, 2015.
[13] Maxim Integrated, MAX30105, [Online]. Available:
https://datasheets.maximintegrated.com/en/ds/MAX30105.pdf
[14] Nordic Semiconductor, NRF52832, [Online]. Available:
https://infocenter.nordicsemi.com/index.jsp?
topic=%2Fcom.nordic.infocenter.nrf52832.ps.v1.1%2Fblockdiagram.html
[15] 維基百科, 人工智慧, [Online]. Available:
https://zh.wikipedia.org/wiki/%E4%BA%BA%E5%B7%A5%E6%99%BA%E8%83%BD
[16] Alan Davis Babu, Artificial Intelligence vs Machine Learning vs Deep
Learning (AI vs ML vs DL), [Online]. Available:
https://medium.com/@alanb_73111/artificial-intelligence-vs-machine-learning-
vs-deep-learning-ai-vs-ml-vs-dl-e6afb7177436
[17] 維基百科, 機器學習, [Online]. Available:
https://zh.wikipedia.org/wiki/%E6%9C%BA%E5%99%A8%E5%AD%A6%E4%B9%A0
[18] OOSGA, 機器學習-定義、演算法、以及商業應用,
[Online]. Available: https://oosga.com/machine-learning/
[19] V. Zocca, G. Spacagna, D. Slater, and P. Roelants. Python Deep learning.
Birmingham, UK:Packt Publishing, 2017.
[20] Jagreet Kaur Gill, Automatic Log Analysis using Deep Learning and AI,
Xenonstack, 27 Aug, 2020. [Online]. Available:
https://www.xenonstack.com/blog/log-analytics-deep-machine-learning/
[21] 盧家鋒, 濾波器設計與應用, [Online]. Available:
https://www.ym.edu.tw/~cflu/MedSigProcess_Class07_CFLu.pdf
[22] Andrew W. Moore, Regression and Classification with Neural Networks,
[Online]. Available:
https://www.cs.bgu.ac.il/~elhadad/nlp21/moore/neural.pdf
[23] Yeh James, 卷積神經網絡介紹,[Online]. Available:
https://medium.com/jameslearningnote
[24] Nils, Introduction to 1D Convolutional Neural Networks in Keras for Time
Sequences, Medium. 2018. [Online]. Available:
https://blog.goodaudience.com/introduction-to-1d-convolutional-neural-
networks-in-keras-for-time-sequences-3a7ff801a2cf
[25] IEEE Standards Association, IEEE Standard for Wearable Cuffless Blood
Pressure Measuring Devices, IEEE Standard, 2014.
[26] P. C.-P. Chao, C.-C. Wu, D. H. Nguyen, B.-S. Nguyen, P.-C. Huang, and V.-H.
Le, “The machine learnings leading the cuffless PPG blood pressure sensors
into the next stage,” IEEE Sensors Journal, Apr. 2021.