| Summary: | 碩士 === 國立臺北科技大學 === 電機工程系所 === 93 === Recently the wavelet-based electrocardiogram (ECG) compression has been widely used for storage and transmission in order to decrease the drawback of large amount of original ECG database. The Percent of Root-mean-square Difference (PRD) distortion measure is one of the most famous criteria in ECG compression. It is a simple mathematical distortion measure, but shorts of diagnostic information. In the thesis we use another criterion called Weighted Diagnostic Distortion (WDD) instead. The WDD can keep track of most important diagnostic information at the encoding stage, and has been proved to be superior to other measures in the Mean Opinion Score test.
In this study, we use the MIT-BIH Arrhythmia Database which consists of a lot of various-type ECG signals to carry out the simulation. At first, the well-known 9/7 Daubechies filter is used to implement the wavelet transform. Then, vector quantization (VQ) is adopted to quantize the wavelet coefficients and produce the error vector. Finally, the SPIHT algorithm, an efficient coding strategy, is exploited in coding the error vector and sorts a list of top n candidates for choice. As usual we look for and pick out the best candidate with the lowest WDD to reconstruct the ECG signal.
|
|---|
