Measure of regularity in discrete time signals

Discrete time series can be treated through linear or non-linear mathematical procedures in order to find specific properties. Last decades, non-linear analysis methods brought valuable results in discrete time series analysis and prediction. Statistical signal processing methods as entropy measures have become important tools in the analysis of time series data, especially in physiology and medicine. Generally, entropy measures the degree of regularity in systems and usually it should be able to quantify the complexity of any underlying structure in the discrete time signals. This paper proposes approximate entropy and sample entropy calculations on synthesized test signals with specific properties in order to try to find hidden properties. Approximate entropy and sample entropy being mathematical algorithms created to measure the regularity or predictability within a time series, are extremely sensitive to their used parameters as length of the data segment and length of data.