Math anxiety (MA), i.e. a trait factor describing the feelings of tension, apprehension, and fear during mathematics-related situations, has attracted increasing interest in recent years, due to its importance in people's daily life and career development, especially in our modern digital world. Although the measurement of individuals' trait MA has mostly relied on self-reported psychological scales, emerging studies are seeking for objective measurement by using behavioral or neurophysiological data. The present study, for the first time, investigated the neurophysiological signatures of trait MA in a cohort of high school students during their 90-minute real final-term math exam. Wrist-worn wearable devices were used for recording their autonomic nervous system activities, including skin conductance (SC) and heart rate (HR). The calculation of pairwise correlation revealed that both SC and HR could reflect the individual's math evaluation anxiety (MEA) score, which is one of the two sub-components of trait MA. Specifically, the tonic level of SC was negatively correlated with MEA during the 5-minute pre-exam period when the students were anticipating the exam, whereas HR was positively correlated with MEA at two later time windows during the exam (63-65 minutes and 82-85 minutes, after the start of the exam). A leave-one-out regression analysis revealed a correlation (r = .349, p = .094) between the self-reported MEA scores and the scores predicted by these neurophysiological signatures. Our findings provide neurophysiological evidence of trait MA in a real-life context and demonstrate the potential of implementing an objective measurement of trait MA based on neurophysiological signals.