Privacy Enhanced Speech Emotion Communication using Deep Learning Aided Edge Computing
This addresses privacy concerns for users in emotion-sensing applications, presenting a novel framework but with incremental technical contributions.
The paper tackles the problem of privacy risks in speech emotion sensing by proposing an adversarial learning framework deployed at the edge to remove private information from speech representations, showing it can hide demographic details and improve emotion identification robustness without significant performance loss.
Speech emotion sensing in communication networks has a wide range of applications in real life. In these applications, voice data are transmitted from the user to the central server for storage, processing, and decision making. However, speech data contain vulnerable information that can be used maliciously without the user's consent by an eavesdropping adversary. In this work, we present a privacy-enhanced emotion communication system for preserving the user personal information in emotion-sensing applications. We propose the use of an adversarial learning framework that can be deployed at the edge to unlearn the users' private information in the speech representations. These privacy-enhanced representations can be transmitted to the central server for decision making. We evaluate the proposed model on multiple speech emotion datasets and show that the proposed model can hide users' specific demographic information and improve the robustness of emotion identification without significantly impacting performance. To the best of our knowledge, this is the first work on a privacy-preserving framework for emotion sensing in the communication network.