The Study Of Multi-Lead ECG Data Compression

Electrocardiogram diagnosis is developing towards remote and intelligent mode with the progress of technology. The remote ECG monitoring system which used technologies of digital signal processing, wire-less communication, and pattern recognition, realize the remote care, monitoring, and diagnoses to the heart-disease patients. In order to facilitate a large number of ECG data storage, analysis and transmission, it must be effectively compressed to improve transfer efficiency and cost savings.This thesis is composed of an ECG compression algorithm based on multiple sparse decompositions and its real-time application on DSP system.The algorithm brought forward by this paper is aiming at promoting more efficiency than single-lead ECG compression, by reducing the multi-lead ECG signals information redundancy integrally. Therefore, the compressible characters of ECG signals have been analyzed deeply. Firstly, the principal component analysis (PCA) is used to quantity the correlation between the V1-V6 chest ECG datas.Then the classic 2-DCT and 2-D wavelet are used to transform the multi-lead signal, and the conclusion is that these methods are not efficiency here. Finally, the novel signal processing method- sparse decomposition is applied to our problem. In this paper, the simultaneous matching pursuit (SMP) is adopted to signal decomposition in over-complete dictionary, while the decomposed coefficients and atomic index numbers will be vector quantized and run-length coded. Then the improved local optimization algorithm (I-SMP) is put forward based on the SMP, making the decomposition coefficients sparser. Simulation results show that the improved method for multi-lead ECG has a good compression.The DSP system, using TMS320VC5509A chip of TI as the kernel, is designed to compress ECG signals in real time with the presented method. The hardware design includes six-way A/D channels design, off-chip memory expansion and Bluetooth communication circuit design. Sampling, compression and wireless transportation of multi-lead ECG signals are accomplished by embedded system programming.