A Model for Synthetic Speech Detection Based on Emotional Variations Using Artificial Intelligence
Today, the automatic recognition of emotions in human speech is gaining increased importance due to the growing role of speech interfaces in human computer interaction applications. In this context, the present study analyzes several methods for detecting synthetic speech and proposes a classification approach based on emotion recognition. Five primary human emotions — anger, boredom, happiness, neutral state, and sadness — are considered, and speech that does not correspond to any of these emotional categories is investigated as potentially synthetic.To achieve more accurate results, statistical methods were employed to combine various feature streams. For emotion detection in speech, a feature vector was constructed by combining the following acoustic features: 16 Linear Predictive Coding (LPC) coefficients, 12 Linear Predictive Cepstral Coefficients (LPCC), 16 Log Frequency Power Coefficients (LFPC), 16 Perceptual Linear Prediction (PLP) coefficients, 20 Mel Frequency Cepstral Coefficients (MFCC), and jitter.The proposed method is based on three classification techniques: Linear Discriminant Analysis (LDA), k-Nearest Neighbors (K-NN), and Hidden Markov Models (HMM). The experimental results demonstrate that the selected features are stable and effective in recognizing emotions and yield high performance based on two evaluation metrics.
[2] Cheng, P. Y., “Automated Recognition
of
Emotion in Mandarin,” MD thesis,
National Cheng Kung University, 2002.
[3] Chuang, Z. J. and C. H. Wu, “Multi
Modal Emotion Recognition from Speech
and Text,” International Journal of
Computational Linguistics and Chinese
Language Processing, 9(2), 2004, pp. 1–18.
[4] Cowie, R., E. Douglas-Cowie, N.
Tsapatsoulis, G. Votsis, S. Kollias, W.
Fellenz and J. G. Taylor, “Emotion
Recognition
Interaction,”
in
Human-Computer
IEEE Signal Processing
Magazine, 18(1), 2001, pp. 32–80.
[5] Davis,
S.
and P. Mermelstein,
“Comparison of Parametric Representations
of
Monosyllabic Word Recognition in
Continuously Spoken Sentences,” IEEE
Transactions on Acoustics, 28(4), 1980, pp.
357–366.
[6] Ekman, P., Handbook of Cognition and
Emotion, John Wiley & Sons, New York,
1999.
[7] Hermansky, H., “Perceptual Linear
Predictive (PLP) Analysis of Speech,Journal of the Acoustical Society of America,
87(4), 1990, pp. 1738–1752.
[8] Holzapfel, H., C. Faion, M. Dennece and
A. Waibel, “Integrating Emotional Cues into
a Framework for Dialogue Management,” in
Proceedings of International Conference on
Multimodal Interfaces, 2002, Pennsylvania,
USA, pp. 141–148.
[9] Iznanoglu, Z. and R. Canel, “Emotive
Alert: HMM-Based Emotion Detection in
Voicemail Messages,” in Proceedings of
Intelligent User Interfaces, 2005, San Diego,
USA, pp. 251–258.
[10] Kaiser, J. F., Discrete-Time Speech
Signal Processing, Prentice Hall, New Jersey,
2002.
[11] Kleinginna, P. R. and A. M.
Kleinginna, “A Categorized List of Emotion
Definitions
with
Suggestions
for
a
Consensual Definition,” Motivation and
Emotion, 5(4), 1981, pp. 345–379.
[12] Kwon, O. W., K. Chan, J. Hao and T.
W. Lee, “Emotion Recognition by Speech
Signals,” in Proceedings of Eurospeech,
2003, Geneva, Switzerland, pp. 125–128.
[13] Le, X. H., G. Ouenot and P. de Lachel,
“Recognizing Emotions from the Audio
Visual Document Using Speech Prosody,” in
Proceedings of the Ninth IEEE International
Symposium
on
Computers
and
Communications, 2004, Alexandria, Egypt,
pp. 580–584.
[14] Linde, Y., A. Buzo and R. M. Gray,
“An Algorithm for Vector Quantizer Design,”
IEEE Transactions on Communications,
28(1), 1980, pp. 84–95.”