Clustering Research on Final Features in Chinese Mandarin under Emotional Differences
Yan Wang, Shaobin Li, Xiaoye Zhang, Fan Zhu · 2020
Under the emotional differences, phoneme feature clustering can unveil the inherent mathematical relationship among the phonemes, laying a research foundation for further emotional analysis using phonemes. In this paper, the audio-visual bimodal phoneme emotional corpus of Chinese Mandarin is firstly constructed with the self-recorded data, and the audio-visual bimodal features of final phoneme are extracted from the samples in the corpus. Then, by using feature clustering with the ISODATA algorithm, the finals of different emotions in the phonetic environment are divided into three and four types respectively. Finally, the clustering results in this paper are compared with the traditional classification methods based on final structural constituents and pronouncing mouth pattern, and the comparison results are further analyzed.