A Dictionary Optimization Method for Reconstruction of ECG Signals after Compressed Sensing

Area of Science:

• Biomedical Engineering
• Signal Processing
• Data Compression

Background:

• Compressed Sensing (CS) is crucial for efficient ECG signal compression and reconstruction.
• Existing data-driven dictionary learning methods can be computationally intensive.
• Optimizing dictionaries tailored to ECG characteristics can enhance performance.

Purpose of the Study:

• To introduce a novel approach for optimizing wavelet dictionaries for ECG signal compression and reconstruction.
• To enhance reconstruction quality by aligning frames to the R-peak for better feature exploitation.
• To evaluate patient-specific optimization algorithms for ECG signal processing.

Main Methods:

• Utilized an overcomplete wavelet dictionary, reduced via a training phase.
• Implemented R-peak alignment for improved dictionary optimization.
• Applied three patient-specific optimization algorithms and tuned their convergence parameters.
• Evaluated performance using Percentage of Root-mean-squared Difference (PRD).

Main Results:

• The optimized dictionary matrix significantly improved ECG signal reconstruction quality compared to other methods.
• Proper tuning of regularization parameters is essential for optimal reconstruction results.
• The Multiple Orthogonal Matching Pursuit (M-OMP) algorithm demonstrated superior performance among the evaluated methods.

Conclusions:

• A patient-specific optimized dictionary approach enhances ECG signal compression and reconstruction.
• Algorithm parameter tuning is critical for achieving state-of-the-art performance.
• M-OMP is the recommended algorithm for this optimized dictionary-based ECG processing system.