SVM-based waist circumference estimation using Kinect.

Area of Science:

• Biomedical engineering
• Medical imaging
• Machine learning applications in healthcare

Background:

• Waist circumference is a critical indicator for diagnosing abdominal obesity and related metabolic diseases.
• Traditional tape measurement for waist circumference requires trained personnel, leading to high costs and time consumption.
• Kinect depth sensors have potential for anthropometric measurements beyond simple distance estimations.

Purpose of the Study:

• To develop and validate a novel method for estimating waist circumference using Kinect depth sensor data.
• To explore the application of Support Vector Machine (SVM) regression for accurate waist circumference measurement.
• To provide a more automated and accessible alternative to manual tape measurements.

Main Methods:

• A dataset was created using Kinect depth images and manual tape measurements from 19 volunteers.
• A Support Vector Machine (SVM) regressor was trained using extracted waist curve vectors from depth images.
• Data augmentation techniques were employed to prevent overfitting, with leave-one-out validation performed on an individual basis.

Main Results:

• The proposed SVM regressor method achieved a mean error of 4.62 cm in waist circumference estimation.
• This error was found to be smaller compared to conventional geometric estimation methods.
• The study demonstrated the feasibility of using depth sensor technology for waist circumference measurement.

Conclusions:

• A viable method for waist circumference measurement using depth sensors has been experimentally validated.
• Future improvements in accuracy are identified, suggesting further potential for this technology.
• This Kinect-based approach can contribute to a patient-centric healthcare model by enabling remote and automated patient monitoring.