現代人有越來越多睡眠相關的問題，睡眠多項生理功能檢查(PSG)是一種耗時的睡眠檢查，它包含很多種生理訊號，像是腦波圖(EEG)、肌電圖(EMG)、心電圖(ECG)，穿戴式裝置提供一個方便且有效率的方法來做長時間的睡眠監測，我們希望可以做出自動睡眠階段分類系統，能應用在穿戴式裝置，並且用較少的取樣數又能維持不錯的準確率。然而，這會面臨到一些挑戰，其中之一是傳送的頻寬有限，我們需要更多頻寬才能傳送從病人身上取得的大量EEG訊號，此外，因為大量的資料傳輸，電池必須提供足夠的能量給設備。

為了解決上述的問題，我們提出一個隱私保護壓縮感知訊號處理(PPCSP)來實現自動睡眠階段分類，我們的資料來源是四個案例來自於四個不同人但同一個EEG通道，我們利用主成分分析(Principal Component Analysis, PCA)的字典(Dictionary)學習來取得資料的特徵。但是主成分分析很依賴資料的型態，每組資料集都需要一個不同的字典，導致省下的頻寬還是很有限，因此，我們針對每個案例，從一個比例平均的資料集訓練出一個通用的字典來取代該案例中所有的字典，它只需要被傳送一次，進而省下頻寬的成本。我們在傳送前利用壓縮感知(Compressed Sensing, CS)降低資料維度，壓縮感知中的壓縮矩陣是一個機率分佈矩陣，對於未來重建資料上可以有保護隱私的效果。

在壓縮資料分析後，我們使用支持向量機(Support Vector Machine, SVM)來做睡眠階段分類，為了解決資料類別不平均的問題，我們透過複製特徵來做資料過度取樣(Oversample)，與其他論文的演算法相比，我們的系統是客製化的，一個案例訓練出一個通用字典。我們用比較少的取樣數達到一樣好的準確率，可以到90%左右，在某些案例中，通用字典的準確率甚至比不同字典的表現還要略高3%左右，壓縮感知結合通用字典不僅有效率地擷取資料特徵，同時也能省下傳送不重要資料的頻寬。

People are getting more and more sleep-relative problems nowadays. Polysomnography (PSG) is a time-consuming record for sleep. It contains many physiological signals, such as electroencephalography (EEG), electromyography (EMG), electrocardiography (ECG). Wireless wearable sensors provide a convenient and efficient way for long-term sleep monitoring. This thesis would like to develop an automatic sleep stage classification system for the implementation of wireless wearable sensors with reduced data from EEG signals with good detection performances. However, there are some challenges such as limited transmission bandwidth and battery power supply.

To solve the aforementioned problems, this work proposed a privacy-preserving compressed signal process (PPCSP) for automatic sleep stage classification. Overnight recorded data of four subjects are included in the experiments. A dictionary was trained with principal component analysis (PCA) to extract features of data. However, PCA is significantly dependent on the data contents, and each dataset needs a different dictionary. So, the saved bandwidth is very limited. Hence, for each subject, a universal dictionary trained from a balanced dataset was used to replace original dictionaries for all data. The dictionary is transmitted only for the first time so that we can lower the cost of transmission bandwidth and reduce the dimension of data with compressed sensing (CS) before transmission. Moreover, the measurement matrix of CS is a random probability matrix that can preserve privacy for reconstruction.

After the compressed data analysis, a support vector machine (SVM) was used to classify sleep stages. Datasets were oversampled by duplicating features to reduce the effect of imbalanced classes. The proposed system was customized and a universal dictionary was trained for a single subject. The experimental results show that our proposed method keeps the accuracy of 90\%, as good as the state-of-the-art works, but with reduced data size. In some subjects, the classification method with a universal dictionary has even higher accuracy than the one with different dictionaries by 3\%. Therefore, the CS combined with a universal dictionary can extract features efficiently and save the bandwidth of data transmission.

[1] C. Iber, S. Ancoli-Isreal, A. C. Jr., and S. Quan. ”The AASM Manual for Scoring of Sleep and Associated Events-Rules: Technical Specification”. Westchester, IL, USA: American Academy of Sleep Medicine, 2007.

[2] Yi Li, Fan Yingle, Li Gu, and Tong Qinye. ”Sleep Stage Classification Based On EEG Hilbert-Huang Transform”. In 2009 4th IEEE Conference on Industrial Electronics and Applications, pages 3676–3681, May 2009.

[3] Farideh Ebrahimi, Mohammad Mikaeili, Edson Estrada, and Homer Nazeran. ”Automatic Sleep Stage Classification Based On EEG signals by Using Neural Networks and Wavelet Packet Coefficients”. In 2008 30th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, pages 1151–1154, August 2008.

[4] Gonzalo M. Rojas, Carolina Alvarez, Carlos E.Montoya, Maria de la Iglesia-Vaya, Jaime E. Cisternas, and Marcelo G´alvez. ”Study of Resting-State Functional Con- nectivity Networks Using EEG Electrodes Position As Seed”. Frontiers in Neuro- science, 66, 2018.

[5] Patrizio Campisi and Daria La Rocca. ”Brain Waves for Automatic Biometric- Based User Recognition”. IEEE Transactions on Information Forensics and Security, 9(5):782–800, 2014.

[6] Khald A. I. Aboalayon and Miad Faezipour. ”Multi-class SVM Based On Sleep Stage Identification Using EEG Signal”. In 2014 IEEE Healthcare Innovation Conference (HIC), pages 181–184, October 2014.

[7] Y. L. Hsu, Y. T. Yang, J. S. Wang, and C. Y. Hsu. ”Automatic Sleep Stage Recurrent Neural Classifier Using Energy Features of EEG Signals”. Neurocomputing, 104:105–114, March 2013.

[8] Aaron Raymond See and Chih-Kuo Liang. ”A Study on Sleep EEG Using Sample Entropy and Power Spectrum Analysis”. In 2011 Defense Science Research Conference and Expo (DSR), pages 1–4, 2011.

[9] Hyungjik Kim and Sunwoong Choi. ”Automatic Sleep Stage Classification Using EEG and EMG Signal”. In 2018 Tenth International Conference on Ubiquitous and Future Networks (ICUFN), pages 207–212, July 2018.

[10] Y. Rachlin and D. Baron. ”The Secrecy of Compressed Sensing Measurements”. IEEE Annual Allerton Conference on Communication, Control, and Computing, pages 813–817, 2008.

[11] Ching-Yao Chou. ”Low-Complexity Compressed Learning for Real-time Wireless Healthcare Monitoring”. Doctoral Dissertation, National Taiwan University Graduate Institute of Electronics Engineering College of Electrical Engineering and Computer Science, June 2019.

[12] Tiziano Bianchi, Valerio Bioglio, and Enrico Magli. ”Analysis of One-Time Random Projections for Privacy Preserving Compressed Sensing”. IEEE Transactions on Information Forensics and Security, 11(2):313–327, 2016.

[13] M.L. Hilton. ”Wavelet And Wavelet Packet Compression of Electrocardiograms”.
IEEE Transactions on Biomedical Engineering, 44(5):394–402, 1997.

[14] Yo-Woei Pua. ”Compressed-domain ECG-based Biometric User Identification”. Master’s Thesis, National Taiwan University Graduate Institute of Electronics Engineering College of Electrical Engineering and Computer Science, June 2020.

[15] S. Dasgupta and A. Gupta. ”An Elementary Proof of The Johnson-Lindenstrauss Lemma”. Random Structures and Algorithms, 22:60–65, June 2002.

[16] Kaihua Zhang, Lei Zhang, and Ming-Hsuan Yang. ”Fast Compressive Tracking”. IEEE Transactions on Pattern Analysis and Machine Intelligence, 36(10):2002– 2015, October 2014.

[17] I. Jolliffe. ”Principal Component Analysis”. 2002.

[18] D.L. Donoho. ”Compressed Sensing”. IEEE Transactions on Information Theory, 52(4):1289–1306, 2006.

[19] M. Unser. ”Sampling-50 Years After Shannon”. Proceedings of the IEEE, 88(4):569– 587, 2000.

[20] Joel A. Tropp and Anna C. Gilbert. ”Signal Recovery From Random Measurements Via Orthogonal Matching Pursuit”. IEEE Transactions on Information Theory, 53(12):4655–4666, 2007.

[21] P. Welch. ”The Use of Fast Fourier Transform For The Estimation of Power Spectra: A Method Based On Time Averaging Over Short, Modified Periodograms”. IEEE Transactions on Audio and Electroacoustics, 15(2):70–73, 1967.

[22] M. Aharon, M. Elad, and A. Bruckstein. ”K-SVD: An algorithm for Designing Overcomplete Dictionaries for Sparse Representation”. IEEE Transactions on Signal Processing, 54(11):4311–4322, 2006.

[23] Ahsan H. Khandoker, Marimuthu Palaniswami, and Chandan K. Karmakar. ”Support Vector Machines for Automated Recognition of Obstructive Sleep Apnea Syn- drome From ECG Recordings”. IEEE Transactions on Information Technology in Biomedicine, 13(1):37–48, 2009.

[24] I. Steinwart, D. Hush, and C. Scovel. ”An Explicit Description of The Reproducing Kernel Hilbert Spaces of Gaussian RBF Kernels”. IEEE Transactions on Information Theory, 52(10):4635–4643, 2006.

[25] De-Gang Chen, Heng-You Wang, and Eric C.C. Tsang. ”Generalized Mercer Theorem and Its Application to Feature Space Related to Indefinite kernels”. In 2008 International Conference on Machine Learning and Cybernetics, volume 2, pages 774–777, 2008.

[26] A. Mathur and G. M. Foody. ”Multiclass and Binary SVM Classification: Implications for Training and Classification Users”. IEEE Geoscience and Remote Sensing Letters, 5(2):241–245, 2008.

[27] Chih-Wei Hsu and Chih-Jen Lin. ”A Comparison of Methods for Multiclass Support Vector Machines”. IEEE Transactions on Neural Networks, 13(2):415–425, 2002.

[28] Jhao-Cheng Wu. ”Sleep Apnea Syndrome Screening by Tri-axial Accelerometer, Oximeter and Phenotype Information”. Master’s Thesis, National Tsing Hua University Graduate Institute of Electrical Engineering, pages 63–64, August 2017.

[29] Jason N. Laska and Richard G. Baraniuk. ”Regime Change: Bit-Depth Versus Measurement-Rate in Compressive Sensing”. IEEE Transactions on Signal Processing, 60(7):3496–3505, 2012.

[30] Xin Yuan and Raziel Haimi-Cohen. ”Image Compression Based on Compressive Sensing: End-to-End Comparison With JPEG”. IEEE Transactions on Multimedia, 22(11):2889–2904, 2020.

[31] Hau-tieng Wu, Ronen Talmon, and Yu-Lun Lo. ”Assess Sleep Stage by Modern Signal Processing Techniques”. IEEE Transactions on Biomedical Engineering, 62(4):1159–1168, 2015.

[32] Physionet. https://physionet.org/about/database/.