I have a data of a 1-D time series data (more than 4000 points), they have fixed spacing, sample at fixed frequency.
My task is to 'extract' the part in this data that match/look similar to my input range of data. In this case, length of the data is fixed, i.e. if I am searching using a set of 10 points data, then it searches and compares for 10 points data only (input length is no more than 30 points).
The criteria is they should look similar, if there is a peak, the searched result should have a peak as well, absolute magnitude and offset is not important.
Currently I am only using a simple correlation function to compare my input and a rolling window. And I am trying to improve it (if appropriate), and I have came across various methods like Distance Correlation, ARIMA, wavelet coherence, Dynamic Time Warping and etc.
My question is, will any of these methods improve what simple correlation can give me? are there any other more appropriate method for my rather 'simple' task?