Wavelet-based ECG data compression optimization with genetic algorithm

With a direct impact on compression performance, optimal quantization scheme is crucial for transformbased ECG data compression. However, traditional optimization schemes derived with signal adaption are commonly inherent with signal dependency and unsuitable for real-time application. In this paper, the variety of arrhythmia ECG signal is utilized for optimizing the quantization scheme of wavelet-based ECG data compression based on a genetic algorithm (GA). The GA search can induce a stationary relationship among the quantization scales of multi-resolution levels. The stationary property facilitates the control of multi-level quantization scales with a single variable. For this aim, a three-dimensional (3-D) curve fitting technique is applied for deriving a quantization scheme with linear distortion characteristic. The linear distortion property can be almost independent of ECG signals and provide fast error control. The compression performance and convergence speed of reconstruction quality maintenance are also evaluated by using the MIT-BIH arrhythmia database.