|
JAIST Repository >
b. 情報科学研究科・情報科学系 >
b11. 会議発表論文・発表資料等 >
b11-1. 会議発表論文・発表資料 >
このアイテムの引用には次の識別子を使用してください:
http://hdl.handle.net/10119/13490
|
| タイトル: | Automatic Speech Emotion Recognition in Chinese Using a Three-layered Model in Dimensional Approach |
| 著者: | Li, Xingfeng
Akagi, Masato |
| 発行日: | 2016-03 |
| 出版者: | 信号処理学会 |
| 誌名: | 2016 RISP International Workshop on Nonlinear Circuits, Communications and Signal Processing (NCSP'16) |
| 開始ページ: | 17 |
| 終了ページ: | 20 |
| 抄録: | In this paper, we improve the speaker independent emotion classification of the CASIA Mandarin emotional speech corpus, which is provided by Chinese-LDC covering four basic emotions, angry, happy, neutral, and sad. We achieve this by restoring the human processing on emotion perception with a three layered model. The three layered model is constructed with acoustic features in the bottom layer, semantic primitives in the middle layer, and emotion dimensions in the top layer. To implement the proposed system, we first investigate the optimal acoustic feature set that is related to each emotion dimension, then mapping these acoustic features to emotion dimensions through the estimated semantic primitives by using Fuzzy Inference System (FIS). In addition, with the highly predicted emotion dimensions, emotional classification procedure is addressed using the knowledge of commonalities and differences of humans emotion perception. The experimental results show that improved estimation performance compared to previous study is furnished. |
| Rights: | Copyright (C) 2016 信号処理学会. Xingfeng Li and Masato Akagi, 2016 RISP International Workshop on Nonlinear Circuits, Communications and Signal Processing (NCSP'16), 2016, 17-20. |
| URI: | http://hdl.handle.net/10119/13490 |
| 資料タイプ: | publisher |
| 出現コレクション: | b11-1. 会議発表論文・発表資料 (Conference Papers)
|
このアイテムのファイル:
| ファイル |
記述 |
サイズ | 形式 |
|---|
| NCSP2016_Xingfeng.pdf | | 595Kb | Adobe PDF | 見る/開く |
|
当システムに保管されているアイテムはすべて著作権により保護されています。
|
