Data Feature Extraction Method of Wearable Sensor Based on Convolutional Neural Network.

Area of Science:

• Wearable technology
• Human activity recognition
• Machine learning for health

Background:

• Wearable devices are crucial for monitoring sports, health, and daily activities.
• Advancements in technology necessitate efficient data analysis methods for wearable sensors.
• Accurate feature extraction is key to understanding human behavior from sensor data.

Purpose of the Study:

• To propose a novel method for feature extraction from wearable sensor data.
• To leverage convolutional neural networks (CNNs) for enhanced human behavior recognition.
• To evaluate the method's effectiveness and generalization capabilities across multiple datasets.

Main Methods:

• Data fusion using the Kalman filter for preliminary state estimation.
• Convolutional Neural Network (CNN) application for human behavior recognition.
• Multi-scale feature extraction while preserving data independence.

Main Results:

• The proposed method effectively extracts relevant feature data from wearable sensors.
• Experimental results on five datasets confirm the method's strong generalization ability.
• The approach successfully maintains data independence during feature extraction.

Conclusions:

• The CNN-based method offers a robust solution for wearable sensor data feature extraction.
• The technique is adaptable to various learning tasks, highlighting its versatility.
• This approach enhances the utility of wearable devices for health and activity monitoring.