Method and System for Signal Separation in Wearable Sensors with Limited Data (with Applications to Transabdominal Fetal Oximetry)

Abstract

Researchers at the University of California, Davis have developed method for separating quasi-periodic mixed-signals using a single data trace, enhancing wearable sensor applications.

Full Description

Deep Harmonic Finesse (DHF) is a method designed to address the challenges of separating quasi-periodic, non-stationary signals from a single mixed signal input. This technology is pivotal for wearable systems where collecting large datasets is impractical. By leveraging deep harmonic neural networks and a pattern alignment method, DHF can isolate target signals from noise and other quasi-periodic phenomena, leveraging prior knowledge of time-frequency patterns. This approach is particularly beneficial in applications such as tissue oximetry and blood glucose monitoring, where accurate signal separation can significantly enhance the reliability of sensor readings.

Applications

• Wearable health monitors for blood glucose levels, tissue oximetry, and more. 
• Non-invasive fetal monitoring and other deep tissue sensing applications. 
• Signal processing solutions where data collection is challenging or impractical.

Features/Benefits

• Enables signal separation using only a single data trace, overcoming the need for large datasets. 
• Addresses the overlap of signal frequencies, a limitation of traditional frequency-based filtering techniques. 
• Incorporates a deep harmonic neural network with pattern alignment for improved signal accuracy. 
• Significantly improves signal-to-distortion ratio and mean squared error compared to existing methods. 
• Enhances the reliability of wearable sensor applications in healthcare, such as non-invasive fetal monitoring. 
• Separates quasi-periodic signals in the presence of frequency overlap and limited data availability. 
• Accurately detects physiological parameters in wearable systems amidst noisy and complex signal environments. 
• Reduces the need for extensive data collection in wearable healthcare monitoring systems.

Patent Status

Patent Pending