臺灣博碩士論文加值系統

心臟是人體非常重要的器官，對心臟病患長期的心電圖記錄是相當重要的。然24小時隨身儲存病患的心電圖，需要花費大量的儲存空間，因此在硬體需求上需要依賴心電圖壓縮技術，使其資料量能夠減少方便儲存，並且保持臨床診斷資訊，因此多種心電圖資料壓縮技術因應而生，其中以Wavelet壓縮可得較大的壓縮效能。
傳統的小波轉換是遞迴式的運算，在有限位元的硬體設計上，會產生「位元成長效應」和「誤差傳播效應」，一維非遞迴式離散週期小波轉換(1-D NRDPWT)和可逆式四捨五入線性轉換(RROLT)，稱為1-D RRO-NRDPWT是新近提出的一種運算方法，可有效抑制位元成長和誤差傳播效應。其在量化上使用非線性量化演算法，此方法的特色是只需一個量化參數QF，就可得到線性的失真度(PRD)-壓縮率(CR)曲線。
本論文是使用ALTERA DE2-70 FPGA CycloneⅡEP2C70F896C6，工作頻率為50MHz，目的是整合1-D RRO-NRDPWT和非線性量化演算，總共使用邏輯單元約47539個。在得到第一筆小波係數後，就開始量化運算，小波轉換與量化將會同時動作，量化在QOD機制下，經實驗在3次內即可達到設定的重建品質。單一導程執行每1024筆所需時間約1.5ms，在考量12導程約需0.018s。對於取樣頻率為1KHz，此系統可以達到Real-Time的需求。

Heart disease diagnosis and patient care rely heavily on long-term recording and ambulatory monitoring of ECG signals. For this purpose, some portable ECG sensing systems associated with wireless data transmission have been developed for real-time ECG signal recording. ECG data compression is crucial for reducing power consumption and improving data transmission efficiency. Many ECG data compression methods have been proposed to achieve a high compression ratio (CR) result and preserve clinical information. Recently, a wavelet-based approach attracted much attention of researchers due to both its simplicity and high-compression performance.
For high octave decomposition, traditional wavelet transform suffers from the bit-length-growth effect and error propagation effect. The 1-D NRDPWT associating the reversible round-off linear transformation (RROLT) theorem is used to minimize the bit-length-growth effect and overcome the error propagation problem. The nonlinear quantization algorithm is conducted with two criteria: determining the quantization scales of all levels with a single variable QF and maintaining the three variables, PRD, CR, and QF, in linear relationship.
The aim of this paper is considers the integration of 1-D RRO-NRDPWT and nonlinear algorithms with ALTERA DE2-70 platform. The operating frequency is 50MHz and the total logic cells are 47,539. To achieve the desired ECG compression quality, this architecture only needs 3 iterations under the novel QOD method for each segment ECG signals. The maximum execution time for compressing single channel ECG signal is 1.5 ms. For considering 12-lead ECG signals, the maximum execution time is about 0.018 s, which can meet the requirement of real-time ECG recording system.

摘要 i
英文摘要 ii
誌謝 iii
目錄 iv
表目錄 vi
圖目錄 vii
第一章 緒論 1
1.1研究背景 1
1.2研究動機與目的 3
1.3研究平台 3
1.4論文架構 5
第二章 一維離散週期性小波轉換理論 6
2.1遞迴式離散週期性小波轉換(DPWT) 6
2.2非遞迴式離散週期性小波轉換(NDPWT) 8
2.3可逆式四捨五入線性轉換系統(RROLT) 15
第三章 心電圖壓縮系統 17
3.1基於1-D RRO-NRDPWT的心電圖壓縮系統 17
3.2非線性量化演算法 19
3.3壓縮系統失真分析 23
3.4壓縮品質控制機制(QOD)應用於ECG壓縮 27
3.5 QOD機制之失真度預測 29
第四章 心電圖壓縮系統硬體整合 32
4.1硬體主架構 32
4.2 1-D RRO-NRDPWT硬體規劃 33
4.2.1轉換矩陣寫入SDRAM 35
4.3量化尺度規劃 37
4.3.1量化尺度讀取 39
4.3.2量化運算處理 40
4.3.3失真度計算處理 43
4.3.4 step預測計算 46
4.4主架構流程狀態機 47
第五章 FPGA模擬結果與分析 51
5.1小波係數軟體模擬 51
5.2小波係數硬體驗證 52
5.3量化驗證 54
第六章 結論 59
參考文獻 60

[1]Daubechies，Ten lectures on wavelet，Ser. No.61 in CBMS-NSF series in Applied Mathematics. Philadelphia, PA: SIAM, 1992.
[2]C.T. Ku , H.S. Wang , K.C. Hung, and Y.S. Hung，A novel ECG data compression method based on non-recursive discrete periodized wavelet transform，IEEE Transaction on Biomedical Engineering, vol. 53, pp.2577-2583, Dec. 2006.
[3]Cheng-Tung Ku，A High Efficient ECG Data Compression System Based On Reversible Round-Off Discrete Wavelet Transform，National Kaohsiung First University of Science and Technology，doctoral dissertation, 2007.
[4]C. T. Ku, H. S. Wang, K. C. Hung, and C.F Tsai，A linear quality control design for high efficient wavelet-based ECG data compression，Computer Methods and Programs in Biomedicine，vol.94,pp.109-117，May 2009.
[5]黃啟仁, “一維可逆式四捨五入非遞迴離散週期性小波轉換之FPGA硬體實現,” 國立高雄第一科技大學電腦與通訊工程系研究所碩士論文, 2009.
[6]陳廷翰, “以FPGA實現一基於可逆四捨五入非遞迴離散週期小波轉換心電圖壓縮系統之量化演算法,” 國立高雄第一科技大學電腦與通訊工程系研究所碩士論文, 2009.
[7]T.C. Wu , K.C. Hung , H.S. Wang，Wavelet-Based ECG Data Comression System With Linear Quality Control Scheme，IEEE Transactions on Biomedical Engineering,vol.57,pp.1399-1409，June 2010.
[8]林灶生，Verilog FPGA 晶片設計，全華2009。
[9]麻省理工學院心律不整心電圖資料庫:
http://www.physionet.org/physiobank/database/mitdb/