| Summary: | 博士 === 國立成功大學 === 電機工程學系碩博士班 === 93 === An electrocardiogram (ECG) is an important physiological signal for heart disease diagnosis. Well trained cardiologists are essential for the analysis of ECG data which is a time consuming task and needs great efforts. For the goal of efficient and convenient processing of ECG, automated and computerized ECG processing has become a major topic of research in the area of biomedical engineering. On the other hand, the storage of ECG data has become more effective because of the computerized processing of ECG. However, storage and transmission limitations have made ECG data compression an important feature for most computerized ECG systems. For example, the Holter systems call for long term storage of multichannel ECG data, with the restriction of small physical size and low power consumption. In addition, the capability of efficient transmission of the stored data is becoming a standard requirement. Such systems require a means of ECG data compression which leads to the conflicting requirements of a high compression ratio (CR) versus good signal fidelity. Besides, an efficient ECG compression technique is also needed for large ECG database, which is a very helpful tool in the evaluation of new automatic ECG processing systems.
Although a great number of ECG compression techniques have appeared in the literature, the search for new methods continues, with the aim of achieving greater compression ratio while preserving the clinical information in the reconstructed signal. By observing the ECG waveforms, a fact can be concluded that the heartbeat signals generally show considerable similarity among adjacent heartbeats, along with short-term correlation among adjacent samples. A compression scheme employing the correlation among adjacent heartbeats can thus further improve the compression efficiency. However, most existing ECG compression techniques did not employ inter-beat correlation. In this dissertation, algorithms utilizing the correlation among adjacent heartbeats are developed to achieve high compression ratio while preserving good reconstructed signal quality.
The first part of this dissertation describes a two-dimensional ECG compression scheme. 1-D ECG signal is segmented and aligned to form a 2-D data array. The 2-D ECG array is then wavelet transformed and its wavelet coefficients are encoded using the modified SPIHT. The highly correlated heartbeat waveforms will result in more centralized wavelet coefficients thus higher compression ratio can be achieved. This scheme can achieve high compression ratio while preserving the fidelity of the reconstructed signal.
The second part of this dissertation is to explore direct ECG compression method without using orthogonal transform such as DCT, DWT, etc.. In the proposed method, vector quantization is used to utilize the inter-beat correlation of ECG signal and an efficient sub-sampling algorithm is adopted to encode the quantization errors. This scheme can achieve the goal of controlling the quality of the reconstructed signal along with high compression ratio.
|
|---|
