Fusion Learning on Multiple-Tag RFID Measurements for Respiratory Rate Monitoring

Area of Science:

• Medical Internet of Things (IoT)
• Wearable Biosensors
• Signal Processing

Background:

• The medical Internet of Things (IoT) requires unobtrusive, passively powered sensors for continuous health monitoring.
• Monitoring infant respiratory activity is crucial for early detection of health issues.
• Existing wearable sensors may face challenges with signal quality and artifact interference.

Purpose of the Study:

• To develop and validate a novel algorithm for monitoring infant respiratory activity using wireless, wearable, and passive knitted smart garment sensors.
• To improve the accuracy of respiratory rate detection by enhancing the signal-to-noise ratio (SNR) of Radio Frequency Identification (RFID) measurements.
• To classify and separate respiratory activity from artifacts using advanced signal processing techniques.

Main Methods:

• Utilized multi-tag RFID measurements from knitted smart garment sensors to capture infant respiratory data.
• Employed fusion learning across multiple features from various RFID tags.
• Developed a Regime Hidden Markov Model (HMM) incorporating higher-order Minkowski and Mahalanobis distance features for signal classification and artifact separation.

Main Results:

• The developed algorithm significantly improved the average Signal to Noise Ratio (SNR) from 17.12 dB to 34.74 dB for respiratory rate detection.
• Demonstrated the effectiveness of higher-order features in enhancing signal strength detection within RFID systems.
• Successfully classified and separated respiratory activity, reducing noise and artifacts in the monitored data.

Conclusions:

• The proposed algorithm enhances the utility of multi-tag RFID measurements for unobtrusive infant respiratory monitoring.
• Higher-order features derived from Minkowski and Mahalanobis distances are effective in improving signal detection in RFID systems.
• The algorithm shows potential for broader applications in machine learning for respiratory data classification and other biomedical signal processing tasks.