InfoShield: Privacy-preserving speech representations for mental health screening

Arxiv CS.CL (NLP/language models)

The method aims to enable scalable depression detection while protecting demographic privacy via information-theoretic optimization. It removes demographic information from speech representations while retaining clinical clues.

speech

Arxiv CS.CL (NLP/language models)·Xueyang WuSiyuan LiuKezhuo YangGuang Ling

6 days ago