在當今先進的醫療技術下，已經有多種方法可以有效地檢測婦女的懷孕情況。然而，這些方法往往各自存在一些缺陷，例如對被檢測者可能造成不適或價格昂貴等問題。心電訊號圖（Electrocardiograph，簡稱ECG）是捕捉人體心臟產生的電信號的一種方法。在懷孕期間，女性體內會發生多種生理和代謝的變化，這些變化可能會在心臟功能上反映出來，進而影響ECG的結果。如果能夠利用ECG作為檢測懷孕的方式，婦女們將能夠更安全、更方便地獲得與懷孕相關的訊息。
本篇論文旨在分析機器學習（Machine Learning）中各種模型對人體心電圖進行分類的表現效果。我們使用穿戴裝置（Wearable device）進行資料收集，以方便且非侵入性的方式獲取人體心電圖數據。我們將數據根據不同的處理方式分成三種類型：圖片、原始訊號和頻率域訊號，並對這些數據進行三種分類任務：性別分類、有無身孕分類和胎兒性別分類。在這項研究中，我們將分析哪種數據和模型在各個分類任務中的效果較佳。

In the current era of advanced medical technology, there are already various methods available for effectively detecting pregnancy in women. However, these methods often have their own drawbacks, such as causing discomfort to the subjects or being expensive. Electrocardiograph (ECG), which captures the electrical signals generated by the human heart, can be used as an alternative approach. During pregnancy, women undergo multiple physiological and metabolic changes that may manifest in cardiac function and thus impact ECG results. If ECG can be utilized as a means of pregnancy detection, women will have a safer and more convenient way to obtain relevant information about their pregnancy.
In this paper, we aims to analyze the performance of various machine learning models in classifying human electrocardiograms (ECG). Data collection is done using wearable devices for convenient and non-invasive detection. The data is processed into three types: images, raw signals, and frequency domain signals. Three classification tasks are performed: gender classification, pregnancy classification, and fetal gender classification. In this study, we will analyze which types of data and models yield better performance for each classification task.

[1] M. Thaler, The only EKG book you'll ever need, Wolters Kluwer Health, seventh edition, 2012.
[2] “圖解傅立葉分析”
https://hackmd.io/@sysprog/fourier-transform (accessed July. 15. 2023)
[3] C. Cortes, V. Vapnik, "Support-vector networks," Maching learning, pp. 273–297,1995.
[4] “小小整理網站 SMALLCOLLATION” https://smallcollation.blogspot.com/2013/07/cardiac-conduction-system.html#gsc.tab=0 (accessed July. 15. 2023)
[5] “Electrocardiogram (ECG): working principle, normal ECG wave, application of ECG”
https://www.onlinebiologynotes.com/electrocardiogram-ecg-working-principle-normal-ecg-wave-application-of-ecg/ (accessed July. 15. 2023)
[6] K. He, X. Zhang, S. Ren, and J. Sun, “Deep Residual Learning for Image Recognition.” arXiv, Dec. 10, 2015. Accessed: May 08, 2023. [Online]. Available: http://arxiv.org/abs/1512.03385
[7] Sepp Hochreiter, and Jurgen Schmidhuber, “Long Short-Term Memory in Recurrent Neural Networks.” Neural Computation, 9(8), 1735-1780.
[8] Dan Li; Jianxin Zhang; Qiang Zhang; Xiaopeng Wei, “Classification of ECG signals based on 1D convolution neural network” 2017 IEEE 19th International Conference on e-Health Networking, Applications and Services (Healthcom), IEEE, 2018.
[9] 廖偉杰, “ECG Classifier Based on Inverse Transformation Algorithm and Support Vector Machines” M.S. thesis, National Taiwan University of Techology and Science, Taipei, Taiwan, 2022.
[10] “Apple Watch Series 5 - 技術規格”
https://support.apple.com/kb/SP808?viewlocale=zh_TW&locale=zh_TW (accessed July. 15. 2023)
[11] “一维数据小波阈值去噪-python实现”
https://zhuanlan.zhihu.com/p/157540476 (accessed July. 15. 2023)
[12] Fazle Karim, Somshubra Majumdar, Houshang Darabi, Shun Chen “LSTM Fully Convolutional Networks for Time Series Classification” IEEE Access ( Volume: 6) pp. 62-69, 2017.
[13] Qingxue Zhang, Dian Zhou, Xuan Zeng “PulsePrint: Single-arm-ECG biometric human identification using deep learning” 2017 IEEE 8th Annual Ubiquitous Computing, Electronics and Mobile Communication Conference (UEMCON), IEEE, 2018.