Abstract: A micro-expression is a brief, involuntary expression that leaks out and reflects an individual’s genuine internal emotional state. Existing research on micro-expressions has primarily focused on micro facial expressions, and the discovery and definition of micro-expressions have been confined to the face. This raises an important question: does a similar form of emotional leakage occur in vocal prosody—that is, do vocal micro-expressions exist? Converging evidence suggests that such vocal micro-expressions do exist. In everyday life, listeners are often able to perceive tension and anxiety in others’ speech; theoretically, suppressed emotions may also leak through the vocal channel. How, then, can these leaked vocal micro-expressions be captured? Although vocal micro-expressions may be difficult to consciously detect through auditory perception alone, they can be objectively quantified using speech emotion recognition techniques by drawing on the physical concept of strain to measure transient emotional information leaked in speech. In addition, a series of psychological experiments can be conducted to analyze the characteristics of these transient emotional signals and apply them to deception detection, thereby providing scientific evidence for the existence of vocal micro-expressions. From the perspective of theoretical construction, the present study examines the connotation and extension, structure, and function of the concept of vocal micro-expressions, with the aim of deepening theoretical understanding of micro-expressions and facilitating their use in deception detection.

Key words: vocal micro-expressions, leakage of emotion, speech emotion recognition, deception detection, machine learning