Real-Time Human Posture Recognition Using Embedded Optical Fiber Sensors and Machine Learning

Smart wearable technology is highly popular in biomedical and kinesiology fields due to its lightweight, flexible, comfortable, and adaptive nature. A smart wearable knee pad was designed using a fiber Bragg grating (FBG) flexible sensor for human posture recognition, combined with machine learning algorithms to enhance recognition accuracy. By placing three FBGs with different central wavelengths in key areas of the knee, posture recognition was realized by observing shifts in the FBG central wavelengths. Although FBG sensors are highly sensitive to small deformations, manual analysis of wavelength shifts introduces errors. Therefore, machine learning algorithms were integrated with the sensor to improve automatic posture recognition accuracy. A voting model combining eXtreme gradient boosting (XGBoost) and random forest (RF) was applied to the sensor. Experimental results showed an accuracy of up to 0.9908, with precision, recall, and $F1$ score reaching 0.9883, 0.9897, and 0.9886, respectively. The area under the receiver operating characteristic (ROC) curve reached 0.9966, demonstrating significant improvement over traditional machine learning models. These results highlight the superior performance of machine learning techniques in human posture recognition combined with FBG sensing technology, providing an effective solution for human posture analysis and virtual reality applications.