現今社會，老年化現象趨於普遍，遠距離醫療照護的相關應用越顯重要，透過無線網路傳輸，在監控病人心臟病理情況的同時，我們透過浮水印的嵌入把病人相關資料，如ID訊息、所在地點，透過心電訊號ECG同時傳回資料庫，在收取檔案時，藉由提取浮水印技術，讀取病人ID訊息，做雙重身分確認，減少檔案存放錯誤。 本論文改善現有量化離散小波的演算法，應用ECG心電訊號浮水印的嵌入與提取，改進的演算法的確能夠降低浮水印嵌入後ECG心電訊號所產生的誤差。經由白噪聲、低頻濾波器、重新取樣等三種攻擊的實驗，同樣是量化嵌入方法下，透過MIT-BIH心律不整資料庫所進行訊號的檢測，探討不同轉換抵抗訊號攻擊的抗性比較。 In nowadays society, we pay much attention to a variety of copyright protection issues and the emphasis on personal information. Digital watermarking technology is the most widely used encryption technology in the field of multi-media and medical information. In this thesis, we redesign a wavelet-based digital watermark to Electrocardiogram (ECG) signals to achieve the purpose of protection of patient rights. Certain hidden data for personnel identification are embedded into specified low-frequency coefficients of its wavelet transform according to different quantization sizes using the quantization approach. In addition, since the proposed watermark scheme is not reversible, we evaluate the impact of watermarking to the PQRST complexes of the ECG signal in which we conclude that the impact is small and the watermarked data can meet the requirement of physiological diagnostics. In order to measure the robustness of our scheme, three attacks including noise, resampling, and low-pass filter are adopted to test the watermarked ECG data. The experimental result verifies that the proposed scheme is slightly robust than DCT- and FFT-implemented quantization watermarking techniques after testing with ECG signals from the MIT-BIH arrhythmia database.
