To improve performance of entropy methods for analyzing physiological signals using a novel symbolic approach

So far, the common symbolic approaches cannot improve performance of complexity algorithms i.e., approximate entropy (ApEn) and sample entropy (SmpEn) to distinguishing randomness and chaotic characteristics within electrocardiogram (ECG) signals, because the approaches not only reflect nonlinear properties within signals but also extract a lot of random component. Our study proposed an encoding symbolic approach (ESA) to keep the inherent nonlinear properties and reduce noise within physiological time series. For the symbolized artificial sequences by the ESA at 100, 500 and 5000 lengths, experiment results show that the values of ApEn and SmpEn, monotonically decrease respectively in order of Gau noise, mixed noise, Logistic map and periodic sequences, whereas values of the two algorithms for the symbolized sequences by the TSA and MSA exhibit fluctuation, indicating performance boost of ApEn and SmpEn on distinguishing randomness and nonlinear chaotic properties within time series with the help of ESA. The experiment results on the MIT/BIH cardiac interbeat interval databases showed the symbolized normal sinus rhythm interval time series by the ESA approach yielded the higher ApEn and SmpEn values than the symbolized congestive heart failure interval time series (p<;0.01), and the results indicated that ESA is help to improve accuracies of the ApEn and SmpEn to measure the inherent chaotic properties of the time series.