Accelerated Variant of an EMD-Based-R Peaks Detection Algorithm Involving FFT-Based Time-Domain Down-Sampling and up-Sampling

Abstract

As most of developed empirical mode decomposition (EMD) based R-peaks detection algorithms consume a considerable time of calculation caused by the large length of the input ECG signal, the design of a new technique that allows the acceleration of such methods becomes necessary. Accordingly, a new variant of an EMD-based strategy for R-peaks localization is presented. The new accelerated variant is constituted of three essential parts. The first step is the length reduction of the input signal by means of the truncation in the Fast Fourier Transform (FFT) domain followed by the application of the inverse FFT guaranteeing a suitable time-domain down-sampling. Consequently, the new input signal of a reduced length preserves all medical information contained initially in the original lengthy signal. The second part is dedicated to identify the QRS complex using EMD-based R-peaks detection. This latter comprises a low-pass filter, Empirical Mode Decomposition (EMD) and the Hilbert transform, Finally, the third phase is the time-domain up-sampling using the FFT, the zero padding and the Inverse Fast Fourier Transform (IFFT) to obtain a resulting processed signal which has the same length as the original signal. Next, as a post-processing step, final R-peaks refined localization is achieved. It is noticeable that the new variant ensures same results, in term of accuracy, as the standard method; however, a significant speed-up ratio of 6.95:1 is reported. Additionally, to more prove the effectiveness of the suggested strategy, it has been applied to accelerate two other efficient algorithms and satisfactory speed up ratios of, 7.20:1 and 4.23:1, respectively have been reached.