On the Possibility of Emotion Recognition While Question-Answering Utterances by Voice Analysis for Junior High School Students

In this paper, we show the results of examining the possibility of emotion analysis based on students' voice analysis. Some research focus on the students' facial expressions to understanding their states in online classes. However, applying the conventional emotion estimation model with facial expressions may be difficult in some cases because facial expressions during class are not clarified and defined. Therefore, we focused on students' voice, which they generally use for communication in online classes. We analyzed the spoken voices of 50 junior high school students using commercially available speech emotion analysis software (InTone, Nemesysco Ltd.) that outputs 14 emotional and five conversation states from the utterance voices. We tasked the junior high school students who participated in the experiment with answering five multiple-choice quizzes orally, as same as they answer questions asked by a teacher in a general class. We found those four emotions correlate with the answer correctness: anticipation
$(\boldsymbol{p} < \mathbf{0.001}, \boldsymbol{r} =\mathbf{0.5})$
, excitement
$(\boldsymbol{p}=\mathbf{0.003}, \boldsymbol{r}=\mathbf{0.4})$
, hesitation
$(\boldsymbol{p}=\mathbf{0.030}, \boldsymbol{r}= \mathbf{0.3})$
, and upset
$(\boldsymbol{p}=\mathbf{0.046}, \boldsymbol{r}=\mathbf{0.3})$
. According to the results, we confirmed the possibility of recognizing students' emotions related to their confidence from students' answer voices in class.