Classification of cardiosynchronous waveforms by projection to a Legendre Polynomial sub-space.

Area of Science:

• Biomedical Engineering
• Signal Processing
• Biometrics

Background:

• Radio Frequency Impedance Interrogation (RFII) is explored for noninvasive hemodynamic monitoring and biometric identification.
• Cardiosynchronous waveforms from RFII offer potential for subject identification in challenging environments.
• Developing robust filtering methods is crucial for extracting unique biometric signatures from RFII signals.

Purpose of the Study:

• To investigate filtering techniques for extracting unique biometric signatures from RFII signals.
• To evaluate the effectiveness of Cepstral analysis for dynamic filter parameter estimation.
• To assess the feasibility of using projected signatures in a Legendre Polynomial subspace for classification.

Main Methods:

• Filtering methods, including Cepstral analysis, were examined for biometric signature extraction.
• Signatures were projected onto a Legendre Polynomial subspace to enhance class separability.
• Support Vector Machine (SVM) and k-Nearest Neighbor (k=3) classifiers were applied to a small dataset.

Main Results:

• Both k-Nearest Neighbor and linear SVM achieved high classification accuracy, ranging from 93% to 100%.
• The Legendre Polynomial subspace projection improved class separability for classification.
• The methods demonstrated successful biometric classification from RFII-generated signals.

Conclusions:

• The study demonstrates encouraging results for RFII-based biometric identification, despite a small sample size.
• Further validation with larger datasets is recommended to refine the process.
• RFII shows potential as a robust technology for biometric identification applications.