Estimation of Online Lecture Quality Using Fundamental Frequency Characteristics ExTracted from Student Utterances

ABSTRACT

The number of online lectures has increased against the backdrop of the COVID-19 pandemic. With the increase in online lectures, more methods of evaluating their quality and improving lecture styles are being developed. We proposed a method of estimating online lecture quality using the SD-F0 values of students’ response utterances. First, we confirmed the effectiveness of the SD-F0 values of students’ response utterances in estimating students’ understanding of lectures. Through identification experiments using an online lecture video database, the precision rate of “Understanding” was found to be 80.6%. This suggests that when the SD-F0 value was high, with a high probability, the student understood the lecture content. Next, we analyzed the relationship between the SD-F0 values of the students’ utterances and lecture quality. We confirmed that during the first thirty minutes of a lecture, when the SD-F0 value was high, the lecture was considered high-quality. If the SD-F0 values of the 10% or 20% utterances in a lecture exceeded a boundary set by the SVM (the boundary between high- and low-quality lectures), the lecture could be regarded as high-quality. The identification rate was greater than 80%. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.