睡眠呼吸中止症是一種在睡眠時停止呼吸的危險疾病，患者在睡眠時會發生反覆呼吸暫停的情況，嚴重時甚至危及生命安全。一次睡眠呼吸中止症事件定義為病患睡眠時發生停止呼吸超過十秒以上。睡眠呼吸中止事件可分為兩種類型 : 阻塞型睡眠呼吸中止症(OSA)事件及中樞型睡眠呼吸中止症(CSA)事件。除了以上兩種類型外，尚有一種病症稱為睡眠淺呼吸(Hypopnea)。淺呼吸為呼吸氣流下降超過50%以上。在醫學上使用睡眠呼吸淺呼吸指標(AHI)來判定病人的嚴重程度。罹患睡眠呼吸中止症的患者會有專注力下降、日間嗜睡、情緒暴躁及夜間覺醒等症狀使得生活品質下降。由於發病時處睡眠時間，常容易讓患者忽略甚至未查覺。目前醫院透過睡眠多項生理檢查(Polysomnography, PSG)診斷睡眠呼吸中止症。但由於該檢測花費昂貴、配戴多條感測器使患者感到不舒適以及環境限制，本研究使用低成本的三軸加速度感測器分別貼黏在左胸以及左腹取得之胸腹起伏訊號搭配PSG中的血氧濃度訊號以簡化睡眠呼吸中止症的檢查。

本研究中，取得三軸加速度感測器的胸腹訊號及血氧濃度訊號後，先利用三軸選擇法(TAA selection method)將胸腹訊號做處理，接著萃取胸腹訊號的振幅比例、頻率比例與共變異數以及血氧濃度中的最小值、最大值、中位數、一次微分後的變異數、中位數與最小值的差異與血氧濃度低於百分之三的比率共十項特徵訊號。最後，利用以上十項特徵訊號放進支持向量機(SVM)做訓練後所得出之分類器配合狀態機(state machine)以分類不同病症，達到所要之目標。

在SVM模擬比較中，使用三軸加速度感測器所取得之胸腹訊號分析模擬結果的準確度為86.8%，僅比使用PSG中胸腹帶所取得之胸腹訊號分析模擬結果少約3%，表示可利用三軸加速度感測器取代胸腹帶。而在判斷病人嚴重程度的準確度上，可達80%。

Sleep apnea syndrome (SAS) is a well-known sleep disorder nowadays.
People suffering SAS cease breathing during sleeping beacuse airway are partially or completely blocked caused by upper respiratory collapse repeatedly.
SAS reduces quality of life by daytime sleepiness, memory declination, sexual dysfunction, and even myocardial infarction or cerebral vascular accident during sleeping.
In general, people hardly notice SAS and need to do overnight sleep test, which is called polysomnography (PSG) in the sleep center of the hospital.
PSG test is a limited environment, unconfortable and expensive examination.
In order to simplify the complex diagnosis, this thesis uses two tri-axial accelerometers (TAA), which are low-cost and small sensors, stick on the left side of thorax and abdomen seperately to sense the thoracic (THO) and abdominal (ABD) movement signals.
This thesis proposed a sleep apnea / hypopnea event detection algorithm to detect the sleep apnea/ hypopnea event by THO, ABD moving signals and oxygen saturation SpO2 signal during sleep.
In the algorithm, using TAA selection method proposed in the thesis to transfer the three-dimensional data captured by TAA into one-dimentional signals seperately called TAA-ABD and TAA-THO and segments the two moving signals into 10-second window to extract features, which are cross-correlation, fundamental frequency and maximum amplitude values. The oxygen saturation signal is segmented into 15-second window and the six features, minimum, maximum, median, variance of the first derivative, difference from the median to the minimum, and the area under dip level of 3\% of oxygen saturation are extracted. Then, putting the ten features into Support Vector Machine (SVM) to construct classifiers and using the classifiers to design a state machine to calculate the number of sleep apnea/ hypopnea events.

Introduction 1
1.1 Sleep Apnea Syndrome . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Research Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Organization of This Thesis . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Medical Diagnosis of Sleep Apnea Syndrome 5
2.1 Polysomnography Examination . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Wireless Sensing Architecture . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Sleep Apnea Syndrome Severity and Sleep Apnea Event Types . . . . . . 12
3 Proposed Respiration Signal Processing Algorithm 19
3.1 Modied Principal Component Analysis Algorithm . . . . . . . . . . . . 19
3.2 Trial-Axial Accelerometer Selection Algorithm . . . . . . . . . . . . . . . 24
3.2.1 Signal Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.2 Candidate Axis Selection . . . . . . . . . . . . . . . . . . . . . . 25
3.2.3 Modication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3 Sleep Apnea and Hypopnea Event Detection Algorithm . . . . . . . . . . 30
3.3.1 Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3.2 Signal Segmentation . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.3.3 Feature Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.3.4 Support Vector Machine Approach to Classier Training . . . . . 40
3.3.5 State Machine for Event Detection and Assessment of Event Detection
Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4 Simulation Result 45
4.1 Thoracic and Abdominal Waveform . . . . . . . . . . . . . . . . . . . . . 45
4.1.1 Normal Event . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.1.2 OSA Event . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.1.3 CSA Event . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.1.4 HYP Event . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2 Support Vector Machine Classication Performance . . . . . . . . . . . . 50
4.3 State Machine Detection Performance . . . . . . . . . . . . . . . . . . . . 53
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5 Conclusion and Future Work 57
5.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

[1] N. S. F. of USA. Why do we need sleep. USA. [Online]. Available:
https://sleepfoundation.org/excessivesleepiness/content/why-do-we-need-sleep
[2] ||. How much sleep do we really need? USA. [Online]. Available:
http://sleepfoundation.org/how-sleep-works/how-much-sleep-do-we-really-need
[3] K. D. FNP. Sleep apnea: Causes, symptoms and treatments. [Online]. Available:
http://www.medicalnewstoday.com/articles/178633.php
[4] HELPGUIDE.ORG. Sleep apnea: Symptoms, causes, treatments, and cures for
sleep apnea. [Online]. Available: http://www.helpguide.org/articles/sleep/sleepapnea.
htm
[5] J. M. Andrew Bates, Martin J. Ling and D. K. Arvind. (2010, July)
Respiratory rate and
ow waveform estimation from tri-axial accelerometer data.
[Online]. Available: https://sleepfoundation.org/excessivesleepiness/content/whydo-
we-need-sleep
[6] G. M. Anmin Jin, Bin Yin, H. Duric, and R. M. Aarts.
(2009, Nov) Performance evaluation of a tri-axial accelerometry-based
respiration monitoring for ambient assisted living. [Online]. Available:
https://sleepfoundation.org/excessivesleepiness/content/why-do-we-need-sleep
[7] C.-W.Wang, rototype of home-monitoring system for patents with sleepdisordered
breathing," July 2014.
8] K.-H. .Yang, \Respiration signal processing with tri-axial accelerometer for homecare
system of sleep apnea syndrome," Master's thesis, National Tsing Hua University,
November 2015.
[9] Y.-Y. .Lin, \Sleep apnea syndrome severity abd sleep apnea event types," Master's
thesis, National Tsing Hua University, May 2014.
[10] H.-T. W. I. Daubechies, J. Lu, Synchrosqueezed Wavelet Transforms: an empirical
mode decomposition-like tool. America: Appl. Comput. Harmon. Anal., 2011.
[11] A. Khandoker, M. Palaniswami, and C. Karmakar, \Support vector machines for
automated recognition of obstructive sleep apnea syndrome from ecg recordings,"
Information Technology in Biomedicine, IEEE Transactions on, vol. 13, no. 1, pp.
37{48, 2009.
[12] H. Al-Angari and A. Sahakian, \Automated recognition of obstructive sleep apnea
syndrome using support vector machine classier," Information Technology in
Biomedicine, IEEE Transactions on, vol. 16, no. 3, pp. 463{468, May 2012.
[13] H.-T.Wu, R. Talmon, and Y.-L. Lo, \Assess sleep stage by modern signal processing
techniques," IEEE Transactions on Biomedical Engineering, p. in press, 2015.
[14] R. Berry, R. Budhiraja, D. Gottlieb, D. Gozal, C. Iber, V. Kapur, C. Marcus,
R. Mehra, S. Parthasarathy, S. Quan, S. Redline, K. Strohl, S. Ward, and M. Tangredi,
Rules for scoring respiratory events in sleep: update of the 2007 AASM Man-
ual for the Scoring of Sleep and Associated Events. America: American Academy
of Sleep Medicine, 2012.