REACTS: Robust Ensemble Approach With Comprehensive Temporal and Spectral Features for TTS Speech Detection
IEEE Access · 2026 · 14 · 90597-90607
DOI: 10.1109/access.2026.3701741 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Komal Mathur und 4 weiteren Autor:innen. Die Arbeit nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Manipal Academy of Higher Education (MAHE), Manipal, India.
Abstract (Originalsprache)
The rapid advancement of AI voice synthesis technologies has led to significant security and authentication concerns, as synthetic speech can be used to impersonate individuals, spread misinformation, or commit fraud. Detecting AI-synthesized speech is crucial for maintaining trust in digital communication. The detection of synthetic speech has become increasingly critical as Text-to-Speech (TTS) technologies advance in naturalness and accessibility. This research introduces a comprehensive framework for distinguishing human from synthetic speech using an extensive set of acoustic features and an ensemble learning approach. The methodology combines prosodic elements (pitch, energy), spectral characteristics (MFCCs, chroma, spectral contrast), and voice quality metrics (jitter, shimmer, harmonics-to-noise ratio) with gradient boosting ensemble models. The feature extraction process carefully captures subtle temporal and spectral differences between human and synthetic speech. The research provides insights into the most discriminative features for synthetic speech detection and demonstrates the effectiveness of ensemble learning in this domain.
Förderung
- Manipal Academy of Higher Education (MAHE), Manipal, India
Zitieren
APA
Mathur, K., Shah, A. G., Chowdhury, P., Balachandra, M., & Pang, S. (2026). REACTS: Robust Ensemble Approach With Comprehensive Temporal and Spectral Features for TTS Speech Detection. IEEE Access, 14, 90597-90607. https://doi.org/10.1109/access.2026.3701741
Vancouver
Mathur K, Shah AG, Chowdhury P, Balachandra M, Pang S. REACTS: Robust Ensemble Approach With Comprehensive Temporal and Spectral Features for TTS Speech Detection. IEEE Access. 2026;14:90597-90607. doi:10.1109/access.2026.3701741
BibTeX
@article{Mathur2026,
author = {Mathur, Komal and Shah, Aditya Goutam and Chowdhury, Parva and Balachandra, Mamatha and Pang, Shaoning},
title = {REACTS: Robust Ensemble Approach With Comprehensive Temporal and Spectral Features for TTS Speech Detection},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {90597--90607},
doi = {10.1109/access.2026.3701741}
}
EUVIMED