Modeling and Reconstructing Textile Sensor Noise: Implications for Wearable Technology.

Area of Science:

• Biomedical Engineering
• Signal Processing

Background:

• Wearable sensors offer non-invasive, continuous electrophysiological monitoring.
• Textile sensors integrated into clothing present challenges due to low signal-to-noise ratio (SNR).
• Accurate processing of textile sensor data is crucial for reliable wearable health technology.

Purpose of the Study:

• To develop a novel technique for classifying textile sensor noise (TSN) in electrocardiogram (ECG) signals.
• To enable the creation of realistic, artifact-laden ECG datasets for algorithm validation.
• To improve the development of signal processing algorithms for wearable textile sensors.

Main Methods:

• Simultaneous recording of ECG signals using textile sensors (waist) and gel electrodes (chest).
• Extraction of TSN by subtracting the chest ECG signal from textile sensor data.
• Classification of TSN into slow and fast categories based on morphological and statistical features.
• Modeling TSN using Linear Prediction Coding (LPC) and reproducing artifacts with Gaussian distribution.

Main Results:

• A novel technique for classifying textile sensor artifacts in ECG signals was developed.
• TSN was successfully classified into slow and fast categories based on distinct features.
• Reproduced TSN, preserving morphological and statistical properties, was generated.
• A textile-like ECG signal dataset was created by adding reproduced TSN to clean ECG signals.

Conclusions:

• The developed method effectively classifies TSN in ECG signals.
• The ability to reproduce TSN facilitates the creation of comprehensive datasets for algorithm development and validation.
• This work contributes to advancing the accuracy and reliability of wearable textile-based biosensing systems.