Human or Machine? A Random Forest–Based Framework for Detecting AI-Generated Speech Using Acoustic Feature Analysis
Keywords:
Audio feature extraction, Deepfake audio, Machine learning, MFCC, Random forest, Voice detectionAbstract
The rapid advancement of AI-based voice synthesis technologies has made it increasingly difficult to distinguish authentic human speech from artificially generated audio, raising concerns around misinformation, fraud, and digital security. This study presents an AI-Based Voice Detection system, Voicelytics, that analyzes uploaded audio or video files and classifies the contained voice as human or AI-generated using a Random Forest classifier trained on extracted acoustic features. The system extracts Mel-Frequency Cepstral Coefficients (MFCC), spectral centroid, spectral rolloff, zero-crossing rate, RMS energy, and pitch from input audio, applies feature scaling, and generates a classification with an associated confidence score through a Flask-based web interface. To rigorously evaluate this approach, a Random Forest classifier was trained and tested on the DEEP-VOICE benchmark dataset — 11,778 balanced samples of real human speech and Retrieval-based Voice Conversion (RVC) deepfakes from eight public figures — using an 80/20 stratified train-test split. On the held-out test set, the classifier achieved 98.73% accuracy, 98.24% precision, 99.24% recall, a 98.73% F1-score, a 1.10% Equal Error Rate (EER), and a ROC-AUC of 0.9993. These results are consistent with, and slightly exceed, the 90–98% accuracy range reported by comparable MFCC-plus-Random-Forest studies in the literature. The system is positioned against prior studies spanning perceptual deep learning, transformer-based, ensemble, benchmark-dataset, and classical feature-based approaches to deepfake speech detection. This work contributes an accessible, locally deployable, interpretable, and empirically validated tool for voice authenticity verification in digital communication, media, and security contexts.
References
A. Chaiwongyen et al., “Deepfake speech detection using perceptual pathological features related to Timbral attributes and deep learning,” Applied Sciences, vol. 16, no. 4, Feb. 2026.
L. Y. Gong and X. J. Li, “Deepfake voice detection: An approach using end-to-end transformer with acoustic feature fusion by cross-attention,” Electronics, vol. 14, no. 10, May 2025.
A. Firc, K. Malinka, and P. Hanáček, “Evaluation framework for deepfake speech detection: a comparative study of state-of-the-art deepfake speech detectors,” Cybersecurity, vol. 8, Aug. 2025.
N. V. Kulangareth, J. Kaufman, J. Oreskovic, and Y. Fossat, “Investigation of deepfake voice detection using speech pause patterns: Algorithm development and validation,” JMIR Biomedical Engineering, vol. 9, 2024.
S. A. Al Ajmi, K. Hayat, A. M. Al Obaidi, N. Kumar, M. S. Najim AL-Din, and B. Magnier, “Faked speech detection with zero prior knowledge,” Discover Applied Sciences, vol. 6, May 2024.
R. Ardila et al., “Common voice: A massively-multilingual speech corpus,” Proceedings of the Twelfth Language Resources and Evaluation Conference, May 2020, pp. 4218–4222.
N. Subramani and D. Rao, “Learning efficient representations for fake speech detection,” Proceedings of the AAAI Conference on Artificial Intelligence, vol. 34, no. 4, pp. 5859–5866, Apr. 2020.
M. Sahidullah, T. Kinnunen, and C. Hanılçı, “A comparison of features for synthetic speech detection,” Proceedings of Interspeech, 2015, pp. 2087–2091.
X. Wang et al., “ASVspoof 2019: A large-scale public database of synthesized, converted and replayed speech,” Computer Speech & Language, vol. 64, Nov. 2020.
J. Yi, C. Wang, J. Tao, X. Zhang, C. Y. Zhang, and Y. Zhao, “Audio deepfake detection: A survey,” arXiv, Aug. 2023.
Z. Khanjani, G. Watson, and V. P. Janeja, “Audio deepfakes: A survey,” Frontiers in Big Data, vol. 5, Jan. 2023.
R. Mubarak, T. Alsboui, O. Alshaikh, I. Inuwa-Dutse, S. Khan and S. Parkinson, “A survey on the detection and impacts of deepfakes in visual, audio, and textual formats,” in IEEE Access, vol. 11, pp. 144497–144529, Dec. 2023.
A. Hamza et al., “Deepfake audio detection via MFCC features using machine learning,” in IEEE Access, vol. 10, pp. 134018–134028, Dec. 2022.
L. Breiman, “Random forests,” Machine learning, vol. 45, pp. 5–32, Oct. 2001.
J. J. Bird and A. Lotfi, “Real-time detection of AI-generated speech for deepfake voice conversion,” arXiv, Aug. 2023.
F. Pedregosa et al., “Scikit-learn: Machine learning in Python,” The Journal of Machine Learning Research, vol. 12, pp. 2825–2830, Nov. 2011.
B. McFee et al., “librosa: Audio and music signal analysis in python, SciPy, Jul. 2015.
