Gait can reveal sleep quality with machine learning models.

Area of Science:

• Biomedical Engineering
• Sleep Science
• Machine Learning

Background:

• Assessing sleep quality is crucial for health, but current methods like questionnaires and polysomnography are often intrusive, costly, or time-consuming.
• There is a need for nonintrusive, inexpensive, and convenient methods to evaluate sleep quality.

Purpose of the Study:

• To investigate the potential of using Kinect sensor-based gait analysis to predict sleep quality.
• To develop and validate machine learning models for sleep quality assessment using gait data.

Main Methods:

• Fifty-nine healthy students participated, providing sleep quality data via the Pittsburgh Sleep Quality Index (PSQI) and gait data using Kinect sensors.
• Gait features were extracted after data preprocessing, and machine learning models were trained to predict PSQI scores.
• Statistical analysis, including t-tests, identified key body joints (e.g., Head, Wrist Left, Hip Left) with significant predictive weight.

Main Results:

• Gaussian processes achieved the highest correlation (0.78, p < 0.001) for overall sleep quality prediction.
• Linear regression yielded the best result for predicting daytime dysfunction (0.51, p < 0.001).
• Specific joints like the Head, Spine Shoulder, and Foot Left were found to be highly influential in predicting sleep quality.

Conclusions:

• Gait patterns captured by Kinect sensors can effectively indicate sleep quality.
• This Kinect-based gait analysis presents a promising, less intrusive, and more ecological method for assessing sleep quality, serving as a valuable supplement to existing approaches.