Symbolization-of-Time-series-and-calculation-of-Entropy

Workflow to Calculate Entropy of a list of Time Series after their symbolization

Normalised Corrected Shannon Entropy is calculated for different time series. The following steps are performed on a Signal or time series ' $T$ ' .

1. Mean is calculated for $T$.
2. A Threshold value $\theta$ is chosen and $T$ is converted into a sequence of 0s and 1s based on threshold.
3. This binary sequence is converted into a word sequence based on a choice of Wordlength $w$. The data points are grouped into words based on a moving window having it's size equal to the word length.
4. The number of possible words are finite and are based on the chosen wordlength.
5. Entropy is calculated for these sequence of words. The quantity calculated here is a modified form of the Shannon Entropy called Normalised Corrected Shannon Entropy.

The details of the method can be obtained here Hussain, L., Aziz, W., Alowibdi, J.S. et al. The method was performed on two datasets (EEG and ECG Signals) as part of a project to distinguish different type of signals using NCSE.