Estimating the degree of conflict in speech by employing Bag-of-Audio-Words and Fisher Vectors
作者：
Highlights：
• We focus on conflict intensity estimation from human speech.
• We test three frame-level features and three utterance-level feature extractions.
• We apply PCA on the frame-level features and achieve up to 30% speed-up with it.
• We achieved the highest correlation coefficient ever reported for a public corpus.
• Yet, our scores fall close to the highest possible ones due to task subjectivity.
摘要
•We focus on conflict intensity estimation from human speech.•We test three frame-level features and three utterance-level feature extractions.•We apply PCA on the frame-level features and achieve up to 30% speed-up with it.•We achieved the highest correlation coefficient ever reported for a public corpus.•Yet, our scores fall close to the highest possible ones due to task subjectivity.
论文关键词：Conflict intensity estimation,Machine learning,Bag-of-Audio-Words,Fisher Vectors
论文评审过程：Received 31 May 2020, Revised 15 July 2021, Accepted 15 May 2022, Available online 28 May 2022, Version of Record 1 June 2022.
论文官网地址：https://doi.org/10.1016/j.eswa.2022.117613
