Related Experiment Video
Updated: Apr 4, 2026
06:37
Author Spotlight: Addressing Technical and Subjective Challenges in Measuring Classroom Attention
Published on: December 15, 2023
5.7K
Advanced feature selection and temporal attention mechanisms with Bi-LSTM classifier for optimizing emotion
Gh Mohmad Dar1,2, Radhakrishnan Delhibabu1,2
1School of Advance Sciences, Vellore Institute of Technology, Vellore, Tamil Nadu, India.
Frontiers in Artificial Intelligence
|April 3, 2026
Summary
This study enhances emotion recognition in Kashmiri speech using optimized features and temporal attention with Long Short-Term Memory (LSTM) networks. The attention-augmented LSTM model achieved 90.2% accuracy, improving speech emotion recognition (SER) for low-resource languages.
Area of Science:
- Computational Linguistics
- Speech Processing
- Artificial Intelligence
Background:
- Emotion recognition in speech is challenging, especially for low-resource languages like Kashmiri.
- Traditional methods often struggle with subtle emotional cues and temporal dynamics.
Purpose of the Study:
- To develop an advanced methodology for enhancing emotion recognition in Kashmiri speech.
- To improve the accuracy and robustness of speech emotion recognition (SER) systems.
Main Methods:
- Optimized feature selection identified key acoustic features (MFCCs, LPC).
- Integrated temporal attention mechanisms into Long Short-Term Memory (LSTM) networks.
- Developed an attention-augmented LSTM model for emotion classification.
Main Results:
- The attention-augmented LSTM model achieved 90.2% accuracy, surpassing the baseline LSTM's 86%.
- Significant improvements were observed in precision, recall, and F1-scores across emotional categories.
- The attention mechanism effectively captured subtle emotional variations and temporal dynamics.
Conclusions:
- Attention-based temporal modeling offers a robust framework for SER in low-resource languages.
- The methodology enhances sensitivity and specificity for emotion recognition systems.
- Findings support future SER deployments in digital human-machine interfaces and multilingual settings.
