PRIVACY-PRESERVING ARTIFICIAL INTELLIGENCE FOR EMPLOYEE DEPRESSION DETECTION

Smart overview of the Invention

The patent application discusses a system designed to detect depression in employees using artificial intelligence, while ensuring user privacy. This system employs a learnable optical model integrated with a multi-task deep network. The model captures images that obscure identity details but retain features necessary for detecting depression. The process involves a feature focus network that extracts and analyzes facial cues related to depression through a progressive learning strategy, which is incrementally trained in stages.

Technical Field

The invention relates to user privacy and health, specifically focusing on systems and methods for monitoring user health while preserving privacy. It targets the detection of health issues, such as depression, without compromising employee privacy. The approach addresses the challenge of maintaining confidentiality while supporting employee wellbeing and productivity.

Background

Employers have a vested interest in monitoring employee health to enhance productivity. However, traditional health monitoring methods often compromise privacy, leading to potential breaches of confidentiality. The need for effective depression detection methods that respect privacy is critical, as mental health issues can significantly impact workplace productivity and environment.

Detailed Description

The system leverages a learnable optical model to capture images that protect user identity while allowing for effective depression detection. A feature focus network incrementally trains on identity obfuscation, emotional feature extraction, and depression-specific feature extraction. A multimodal transformer fuses emotional and depression features to produce an accurate depression score, achieved through an end-to-end joint optimization process.

Implementation

The system includes a privacy-preserving lens that obscures identity features but retains depression-related features. The lens is trained to optimize parameters for effective image capture. The captured image is processed through a deep network and a multimodal transformer to ensure privacy while accurately detecting depression. This approach allows entities to support employee mental health confidentially, enhancing wellbeing and productivity.