查詢結果分析
來源資料
頁籤選單縮合
題 名 | 基於RRO-NRDPWT心電圖資料壓縮法中之線性PRD-CR規劃=Linear PRD-CR Programming in RRO-NRDPWT Based ECG Data Compression |
---|---|
作 者 | 辜正同; | 書刊名 | 慈惠學術專刊 |
卷 期 | 3 2007.10[民96.10] |
頁 次 | 頁112-123 |
分類號 | 312.1 |
關鍵詞 | 心電圖; 資料壓縮; 量化演算法; 非遞迴式小波轉換; 可回復性四捨五入線性轉換; ECG; Data compression; Quantization; Wavelet transform; RROLT; |
語 文 | 中文(Chinese) |
中文摘要 | 心電圖 (ECG) 是一種非侵入式檢測廣泛應用於心臟疾病之臨床診斷。心臟病患的照護與即時醫療十分仰賴長期的心電圖紀錄與即時資料傳輸,為達此目的,通常使用行動式擷取裝置以方便患者的日常生活。為滿足如此龐大資料的儲存與傳輸需求,多種心電圖資料壓縮技術乃因應而生,其中以基於小波轉換 (wavelet-based) 的壓縮方法可得最佳壓縮效能。 傳統的小波轉換係採遞迴式金字塔演算法,我們稱此種分頻方式為一維遞迴式小波轉換 (1-DRDWT)。然而l-DRDWT存在位元長度成長效應 (word-length growth effect) 及誤差傳播 (error propagation) 現象。此兩效應使得我們無法有效地控制量化失真,導致 wavelet-based壓縮法的壓縮效能在高壓縮率時呈現急遽下降現象。為克服此困難,我們 發展出一可逆式一維非遞迴式離散週期性小波轉換(l-DRRO-NRDPWT) 。 1-D RRO-NRDPWT結合了一維非遞迴式離散週期性小波轉換 (l-D NRDPWT) 與可回復性四捨五入線性轉換 (RROLT) 性質。l-DNRDPWT用於阻絕誤差傳播的發生,而RROLT的性質為在可逆的條件下,可獲得最小動態範圍 (MDR) 的全整數小波係數值,並且能提供對於每一頻帶量化誤差所造成的重建失真一個定量的描述。本論文將基於RRO-NRDPWT之心電圖壓縮系統,提出一個單參數控制的非線性量化演算法,此量化 處理可獲得線性的PRD-CR特性,可有效的提升壓縮效能,甚至在高CR的條件下,仍然保持低的PRD值。 |
英文摘要 | Electrocardiogram (ECG) signal analysis is a non-invasive modality widely used for heart disease diagnosis. The cardiac really rely on a long-time electrocardiogram recording and real-time data transmission for their home care services and real-time medication. They need an ambulatory monitoring device for their convenient life. Many ECG data compression methods have been proposed for sufficient the request of the mass data storage and transmission. Recently, the wavelet-based approach has attracted much attention from researchers due to simplicity and high compression performance. Traditional wavelet-based ECG data compression methods applied the recursive pyramid algorithm to perform the l-D discrete wavelet transform (DWT). This DWT process is referred to as the l-D RDWT where six stages of decomposition are usually required for each segment of ECG data. However, the l-D RDWT exist the error propagation and word-length-growth effects. So it couldn’t efficiently control the quantization error caused by these effects. The compression performance of wavelet-based will be decreased rapidly in high CR situation. To overcome this drawback, we developed a reversible round-off l-D non-recursive discrete periodized wavelet transform (1-DRRO-NRDPWT). The l-D RRO-NRDPWT, associates the l-D NRDPWT and the reversible round-off linear transformation (RROL T) theorem, is derived for overall stages decomposition that obtains a DC value in the termination level and the wavelet coefficients of all octaves. The l-D NRDPWT can stop the error propagation and the RROLT can get integral coefficients with minimum dynamic range (MDR) and also analysis the influences of reconstructed distortion by each sub-band’s quantized error. In this paper we proposed an ECG data compression method based on the RRO-NRDPWT. We developed a non-linear quantization algorithm that is controlled by a single parameter. The linear relationship between PRD and CR is obtained by this quantization process. The compression performance can be improved efficiently and the PRD is kept at small value even at high CR situation. |
本系統中英文摘要資訊取自各篇刊載內容。