[Analysis of Bioelectrical Impedance for Identification]

Area of Science:

• Biomedical Engineering
• Pattern Recognition
• Biometrics

Background:

• Personal identification is crucial in various applications.
• Existing methods may have limitations in accessibility or cost.
• Bioelectrical impedance analysis (BIA) offers a non-invasive measurement approach.

Purpose of the Study:

• To investigate the feasibility of using bioelectrical impedance spectrum analysis for individual identification.
• To develop and evaluate a pattern recognition model for accurate person recognition.
• To establish a simple and supplementary method for human identification.

Main Methods:

• Measured bioelectrical impedance spectra in hands across a 1-100 kHz frequency range using an AD5933 chip.
• Calculated segmentation slopes of the impedance spectra as characteristic parameters.
• Employed a 3-layer backpropagation neural network, trained and tested on divided datasets.

Main Results:

• The backpropagation neural network model demonstrated effective identification of individuals.
• Achieved high recognition rates: 97.62% for training sets, 88.79% for validation sets, and 86.34% for test sets.
• Overall synthetic recognition rate reached 94.22%, indicating reliable identification of both known and unknown individuals.

Conclusions:

• Bioelectrical impedance analysis combined with pattern recognition is a feasible and reliable method for personal identification.
• The developed neural network model shows strong potential for accurate and efficient individual recognition.
• This approach offers a simple, non-invasive, and supplementary tool for human identification.