Answers are grounded in this paper and cite their sources.
Emotion recognition systems frequently depend on sensitive personal data, such as facial expressions, voice patterns, and textual input, making privacy protection a critical concern. This review provides a comprehensive analysis of emotion-preserving anonymization methods, focusing on their application to emotion classification tasks. Using the PRISMA methodology, a systematic literature review was conducted across multiple databases, including Google Scholar, PubMED, Medline, Embase, PMC, NCBI Bookshelf, and NIH Manuscripts. Searches spanned a decade (2014–2024) and employed queries such as “anonymization + emotion,” “privacy + preserve + emotion,” and “de-identification + emotion,” yielding 49,700 initial results. After removing duplicates and manual screening, 70 publications were selected for in-depth analysis. The study categorizes de-identification techniques, emphasizing their effectiveness in maintaining emotional cues while safeguarding individual identities across multi-modal datasets. A systematic evaluation of these methods reveals trade-offs between privacy protection and data utility, with particular attention to the accuracy of emotion classification after anonymization. The review identifies gaps in current approaches, including the need for advanced techniques optimized for multi-modal datasets, cross-cultural robustness, and real-world applicability in privacy-sensitive domains such as mental health or education. This review emphasizes the importance of advancing privacy-preserving anonymization techniques, which enables ethical emotion recognition research, while preserving the privacy of individuals.
No takes yet. Share an insight, caveat, or question.
Osuchowski et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: