Curve aligning approach for gait authentication based on a wearable accelerometer.

Area of Science:

• Biometrics
• Wearable technology
• Signal processing

Background:

• Gait authentication using wearable accelerometers is a novel biometric.
• Matching gait patterns is crucial for authentication performance.
• Existing methods like correlation and DTW have limitations.

Purpose of the Study:

• To introduce curve aligning as a new method for gait pattern matching.
• To compare curve aligning with correlation and DTW.
• To fuse pattern-matching methods using a support vector machine (SVM) for enhanced authentication.

Main Methods:

• Collected acceleration data from the ankles of 22 walking subjects.
• Implemented curve aligning, correlation, and dynamic time warping (DTW) for gait pattern matching.
• Utilized a support vector machine (SVM) for decision-level fusion of pattern-matching results.

Main Results:

• The fusion of curve aligning with backward-forward accelerations and DTW with vertical accelerations substantially improved authentication performance.
• The proposed fusion algorithm achieved a mean equal error rate (EER) of 0.794% with a standard deviation of 0.696%.
• This fusion method significantly outperformed non-fusion algorithms, where the best achieved an EER of 3.03%.

Conclusions:

• Curve aligning is an effective new method for gait pattern matching in biometric authentication.
• Fusion of multiple pattern-matching techniques, particularly curve aligning and DTW, significantly enhances gait authentication accuracy.
• The proposed fusion algorithm demonstrates high potential for identity identification and other applications of gait analysis.