On the Emotion Understanding of Synthesized Speech
This work highlights a key limitation in evaluating emotional expressiveness in speech synthesis, impacting researchers and developers in voice interaction and AI.
The study critically examined the assumption that emotion understanding models transfer to synthesized speech, finding that current Speech Emotion Recognition (SER) models fail to generalize to synthesized speech due to representation mismatches and that generative models rely on textual semantics over paralinguistic cues.
Emotion is a core paralinguistic feature in voice interaction. It is widely believed that emotion understanding models learn fundamental representations that transfer to synthesized speech, making emotion understanding results a plausible reward or evaluation metric for assessing emotional expressiveness in speech synthesis. In this work, we critically examine this assumption by systematically evaluating Speech Emotion Recognition (SER) on synthesized speech across datasets, discriminative and generative SER models, and diverse synthesis models. We find that current SER models can not generalize to synthesized speech, largely because speech token prediction during synthesis induces a representation mismatch between synthesized and human speech. Moreover, generative Speech Language Models (SLMs) tend to infer emotion from textual semantics while ignoring paralinguistic cues. Overall, our findings suggest that existing SER models often exploit non-robust shortcuts rather than capturing fundamental features, and paralinguistic understanding in SLMs remains challenging.