An efficient strategy for evaluating similarity between time series based on Wavelet / Karhunen-Loève transforms

Area of Science:

• Biomedical Engineering
• Signal Processing
• Health Informatics

Background:

• Physiological time series analysis is crucial for disease monitoring.
• Efficient similarity measures are needed for complex biomedical data.
• Telemonitoring generates large volumes of physiological data requiring robust analysis.

Purpose of the Study:

• To develop an innovative and efficient similarity measure for physiological time series.
• To combine Haar wavelet decomposition and Karhunen-Loève transform for signal analysis.
• To reduce computational complexity for clinical and telemonitoring applications.

Main Methods:

• Utilized Haar wavelet decomposition to represent signals using orthogonal basis.
• Applied Karhunen-Loève transform for optimal reduction of basis sets.
• Calculated similarity via Euclidean distance on linear combination coefficients, using an iterative scheme.

Main Results:

• Developed a computationally efficient similarity measure for physiological time series.
• Successfully validated the method within the HeartCycle project.
• Applied the technique to blood pressure signals for hypertension episode recognition.

Conclusions:

• The proposed method offers a simple and fast approach for physiological time series similarity evaluation.
• Its efficiency makes it suitable for computationally demanding clinical applications, including telemonitoring.
• The validated strategy aids in recognizing hypertension episodes using telemonitoring data.