Stack LSTM-Based User Identification Using Smart Shoes with Accelerometer Data.

Area of Science:

• Biometrics
• Machine Learning
• Wearable Technology

Background:

• Traditional user identification methods using gait analysis often require pre-processing to segment individual steps.
• Partial or incomplete gait data can pose challenges for classification accuracy in existing approaches.

Purpose of the Study:

• To develop a novel user identification method using accelerometer data from smart shoes.
• To overcome limitations of step-division pre-processing in human walking data analysis.

Main Methods:

• A stacked long short-term memory (LSTM) network was designed for direct user identification from accelerometer data.
• The LSTM network was trained on walking data of random sizes and locations to learn from partial data.
• No additional analysis, such as step division, was required before classification.

Main Results:

• The proposed LSTM-based method achieved high average recognition rates: 93.41% (2.6s), 97.19% (3.9s), and 98.26% (5.2s) using 10m walking data.
• The system demonstrated effective user classification without the need for complex pre-processing steps.

Conclusions:

• The developed LSTM-based method provides an effective and efficient approach for user identification using smart shoe accelerometer data.
• This method successfully handles partial data, offering a robust solution for biometric authentication in wearable devices.