Related Experiment Video
Updated: Jan 16, 2026
05:48
Author Spotlight: Investigating the Impact of Emotional Prosodies on Voice Recognition and Perception
Published on: August 9, 2024
2.0K
TSFNet: A Temporal-Spectral Fusion Network for advanced speech emotion recognition in medical applications
Xinran Li1, Peilin Huang1, Xiaojiang Peng1
1School of Artificial Intelligence, Shenzhen Technology University, 3002 Lantian Road, Shenzhen, 518118, Guangdong, China.
Artificial Intelligence in Medicine
|October 5, 2025
Summary
This study introduces TSFNet, a novel Temporal-Spectral Fusion Network, for advanced speech emotion recognition (SER). TSFNet effectively captures emotional nuances in speech, showing great potential for medical diagnostics and patient care.
Area of Science:
- Artificial Intelligence
- Speech Processing
- Computational Linguistics
Background:
- Speech emotion recognition (SER) is crucial for human-machine interaction and medical applications.
- Existing SER methods struggle to capture subtle emotional nuances vital for medical diagnostics.
- Accurate SER can significantly improve patient care and monitoring systems.
Purpose of the Study:
- To introduce TSFNet, a Temporal-Spectral Fusion Network, for enhanced SER.
- To effectively integrate temporal and spectral speech features for nuanced emotion detection.
- To leverage large-scale pre-trained models for improved temporal characteristic extraction in speech.
Main Methods:
- Developed TSFNet, a Temporal-Spectral Fusion Network.
- Integrated temporal and spectral speech features using a plug-and-play pre-trained model.
- Evaluated TSFNet performance on six public speech emotion datasets.
Main Results:
- TSFNet significantly outperformed existing SER baselines across multiple datasets.
- Achieved high unweighted accuracies: 95.57% (Savee), 92.67% (Crema-D), 85.71% (IEMOCAP), 100.00% (Tess), 95.86% (Emovo), and 80.43% (Meld).
- Demonstrated TSFNet's capability in capturing complex emotional details.
Conclusions:
- TSFNet offers a highly efficient approach to SER by fusing temporal and spectral information.
- The network shows significant potential for advancing medical diagnostic tools.
- TSFNet can enhance patient monitoring systems through accurate emotion recognition.
