Most recent post:
25 Jan 2022
Suppose we are given two discrete signals corresponding to the speech resulting from pronouncing a given sentence. They might have been pronounced by different people and in different intonations and speed, and we would like determine the similarity between these signals.
In this post we’ll discuss dynamic time warping, a technique that can be used to compare the similarity between time series allowing for small variances in the time scales. We’ll provide Python implementations of an exact algorithm $O(N^2)$ based on dynamic programming and an approximated one that runs in $O(N)$.