# Optical Sensor-Based Approaches in Obesity Detection: A Literature Review of Gait Analysis, Pose Estimation, and Human Voxel Modeling

## TL;DR

This paper reviews how optical sensors and AI can detect obesity through gait and posture analysis, offering non-contact alternatives to traditional methods.

## Contribution

The paper introduces hybrid sensor approaches and AI-driven frameworks for dynamic obesity detection, addressing scalability and ethical issues.

## Key findings

- Optical sensors like OpenPose and MediaPipe enable real-time, non-contact obesity detection through gait and posture analysis.
- Hybrid sensor systems improve robustness in uncontrolled environments and diverse populations.

## Abstract

What are the main findings?
This review examines optical and vision-based sensors including pose estimation (OpenPose, MediaPipe), infrared depth sensing, and 3D body modelling for non-contact obesity detection through gait and posture analysis.It highlights AI-driven, real-time capabilities and addresses challenges such as measurement accuracy, environmental factors, scalability, and ethical concerns (privacy, consent, algorithmic bias). Hybrid sensor approaches are proposed to improve robustness.

This review examines optical and vision-based sensors including pose estimation (OpenPose, MediaPipe), infrared depth sensing, and 3D body modelling for non-contact obesity detection through gait and posture analysis.

It highlights AI-driven, real-time capabilities and addresses challenges such as measurement accuracy, environmental factors, scalability, and ethical concerns (privacy, consent, algorithmic bias). Hybrid sensor approaches are proposed to improve robustness.

What is the implication of the main finding?
The findings show strong potential of AI-driven, contactless sensors to improve obesity screening and personalized monitoring beyond traditional static methods.Successful translation into practice requires addressing technical and ethical issues to ensure equitable, reliable, and scalable adoption in healthcare and public health.

The findings show strong potential of AI-driven, contactless sensors to improve obesity screening and personalized monitoring beyond traditional static methods.

Successful translation into practice requires addressing technical and ethical issues to ensure equitable, reliable, and scalable adoption in healthcare and public health.

Optical sensor technologies are reshaping obesity detection by enabling non-invasive, dynamic analysis of biomechanical and morphological biomarkers. This review synthesizes recent advances in three key areas: optical gait analysis, vision-based pose estimation, and depth-sensing voxel modeling. Gait analysis leverages optical sensor arrays and video systems to identify obesity-specific deviations, such as reduced stride length and asymmetric movement patterns. Pose estimation algorithms—including markerless frameworks like OpenPose and MediaPipe—track kinematic patterns indicative of postural imbalance and altered locomotor control. Human voxel modeling reconstructs 3D body composition metrics, such as waist–hip ratio, through infrared-depth sensing, offering precise, contactless anthropometry. Despite their potential, challenges persist in sensor robustness under uncontrolled environments, algorithmic biases in diverse populations, and scalability for widespread deployment in existing health workflows. Emerging solutions such as federated learning and edge computing aim to address these limitations by enabling multimodal data harmonization and portable, real-time analytics. Future priorities involve standardizing validation protocols to ensure reproducibility, optimizing cost-efficacy for scalable deployment, and integrating optical systems with wearable technologies for holistic health monitoring. By shifting obesity diagnostics from static metrics to dynamic, multidimensional profiling, optical sensing paves the way for scalable public health interventions and personalized care strategies.