A novel method for fast Change-Point detection on simulated time series and electrocardiogram data

Area of Science:

• Signal Processing
• Biomedical Engineering
• Data Analysis

Background:

• The Kolmogorov-Smirnov (KS) statistic is a common method for change point detection.
• However, the KS statistic has limitations, including being time-consuming and sometimes inaccurate for abrupt change point (CP) problems.

Purpose of the Study:

• To develop a novel, fast, and efficient method for detecting abrupt change points in time series data.
• To evaluate the performance of the proposed method against existing techniques.

Main Methods:

• A new Haar Wavelet and KS statistic (HWKS) method was developed.
• This involved constructing Binary Search Trees (BSTs) using multi-level Haar Wavelet transforms.
• A modified KS statistic and BST-based search rules were implemented for fast CP detection.

Main Results:

• Simulated time series data demonstrated that the HWKS method is faster, more sensitive, and more efficient than KS, HW, and T methods.
• Application to electrocardiogram (ECG) time series showed HWKS can quickly and efficiently identify abrupt changes in segments with high data fluctuation.

Conclusions:

• The proposed HWKS method offers a significant improvement for fast and efficient abrupt change point detection in time series.
• HWKS is particularly useful for analyzing biomedical signals like ECGs, aiding in the diagnosis of health states.