Multi-Stream Convolution-Recurrent Neural Networks Based on Attention Mechanism Fusion for Speech Emotion Recognition

Multi-Stream Convolution-Recurrent Neural Networks Based on Attention Mechanism Fusion for Speech Emotion Recognition

The quality of feature extraction plays a significant role in the performance of speech emotion recognition. In order to extract discriminative, affect-salient features from speech signals and then improve the performance of speech emotion recognition, in this paper, a multi-stream convolution-recur...

Full description

Bibliographic Details
Main Authors: Huawei Tao, Lei Geng, Shuai Shan, Jingchao Mai, Hongliang Fu
Format: Article
Language:English
Published: MDPI AG 2022-07-01
Series:Entropy
Subjects:
speech emotion recognition
feature extraction
hybrid neural network
multi-head attention mechanism
feature fusion
Online Access:https://www.mdpi.com/1099-4300/24/8/1025

Internet

https://www.mdpi.com/1099-4300/24/8/1025

Similar Items