Emotion prosody perception requires simultaneous processing of acoustic cues and speaker identity. While listeners effortlessly decode natural speech, AI synthetic voices introduce cognitive complexities due to subtle acoustic atypicalities. It remains unclear how these synthetic features interact with a listener's prior social knowledge and memory of a familiar speaker. This study investigated how speech sources (human vs. AI) and speaker familiarity affect emotion recognition accuracy and cognitive load. A within-subject task with Mandarin-speaking adults evaluated behavioral (accuracy, reaction time) and physiological data (heart rate variability). Results showed that human voices yielded significantly higher accuracy and faster processing times than AI voices, while HRV did not significantly differentiate between conditions. These findings show that decoding synthetic speech is gated by top-down social cognition, highlighting limitations in current AI synthesis technologies.
Recent advances in emotional voice conversion (EVC) have enabled the generation of expressive synthetic speech, raising new concerns in audio deepfake detection. Existing approaches treat speech as a homogeneous signal and largely overlook its internal phonetic structure, limiting their interpretability in emotionally conditioned settings. In this work, we propose a phoneme-level framework to analyze emotionally manipulated synthetic speech using real and EVC-generated speech under matched emotional conditions with shared transcripts, phoneme-aligned TextGrids, and WavLM-based embeddings. Our results show that phoneme behavior varies across categories, with complex vowels and fricatives exhibiting higher divergence while simpler phonemes remain more stable. Phonemes with larger distributional differences are also found to be more easily detected, consistently across multiple emotions and synthesis systems. These findings demonstrate that phoneme-level analysis is an effective and interpretable approach for detecting emotionally manipulated synthetic speech.
Vamshi Nallaguntla, Shruti Kshirsagar, Anderson R. Avila
Wichita State University, Wichita, KS, USA. · Institut national de la recherche scientifique (INRS–EMT), QC, Canada · INRS-UQO Mixed Research Unit on Cybersecurity, Gatineau, Canada
This paper explains the principles and provides examples of a new method for distinguishing between FAKE human speech synthesized by generative AI and natural speech. Since synthetic speech is generated based on information from a limited set of training spectra, the variety of vowels - which are key to identifying individuals - is limited. In contrast, natural speech exhibits a more diverse distribution of vowel spectra due to the flexibility of the human articulatory organ. In this paper, using Japanese - a Syllabary limited to five vowel phonemes, each of which corresponds one-to-one with a specific sound - as an example, we outline a method for distinguishing between synthetic and natural speech reading the same text by analyzing the spectral distributions. If we normalize the spectra of speech sounds and regard them as probability density functions for the frequency bands received by the hair cells of the human cochlea, and evaluate the distance between spectra using the Wasserstein metric, the Wasserstein distances between the vowels of synthetic speech are short. By preserving this distance and performing a topological mapping using persistent homology, the spectral probability density functions of synthetic and natural speech can be decomposed into clusters.
Yusei Tamura, Shigekazu Ishihara, Ken Ito
Graduate School of Interdisciplinary Information Studies The University of Tokyo · Department of Psychology, Faculty of Health and Wellness Sciences Hiroshima International University · Interfaculty Initiative in Information Studies The University of Tokyo
Speech conveys information through both words and vocal delivery. We evaluate four leading production realtime voice systems-OpenAI's GPT Realtime 2, Google's Gemini 3.1 Flash Live, and Alibaba's Qwen3.5 Omni Plus and Omni Flash-on tasks where the words and the delivery patterns both convey meaningful information. Across three consequential scenarios, all four systems act on the words rather than the voice. They end calls with crying callers who insist nothing is wrong, approve wire transfers authorized in frightened voices, and enroll callers whose agreement is clearly sarcastic. Surprisingly, this is often not a failure of perception. When asked directly, three of the four systems reliably identify the distress, fear, or sarcasm they later ignore when making decisions. We observe a similar pattern when these realtime voice systems estimate accent and age, as their responses frequently follow the biases of the words rather than the acoustic properties of the speaker. We term this disconnect between perception and action the emotional intelligence gap of voice AI. Prompting systems to explicitly attend to vocal delivery improves performance only partially and inconsistently. Our findings show that current realtime voice AI systems often behave as if speech had been reduced to a transcript, suggesting that they should be used with caution in settings where the tone and emotion of delivery convey important information.