Omslagsbillede

Enhancing speech emotion recognition through parallel CNNs with transformer encoder and co-attention

Guardado en:
Udgivet i:PeerJ Computer Science (Nov 7, 2025)
Hovedforfatter: Hashem, Ahlam
Andre forfattere: Arif, Muhammad, Alghamdi, Manal, Al Ghamdi, Mohammed A, Almotiri, Sultan H
Udgivet:
PeerJ, Inc.
Fag:
Communication
Artificial neural networks
Recognition
Neural networks
User experience
Emotions
Coders
Datasets
Human-computer interaction
Human-computer interface
Computer mediated communication
Emotion recognition
Attention
Classification
Effectiveness
Phonetics
Algorithms
Audio data
Acoustics
Ablation
Interpersonal communication
Speech
Speech recognition
Subjectivity
Human technology relationship
Robustness
Models
Acknowledgment
Generalizability
Machinery
Accuracy
Online adgang:Citation/Abstract
Full Text + Graphics
Tags: Tilføj Tag
Ingen Tags, Vær først til at tagge denne postø!
Resumen:Speech emotion recognition (SER) plays a pivotal role in enabling machines to determine human subjective emotions based only on audio information. This capability is essential for enabling effective communication and enhancing the user experience in human-computer interactions (HCI). Recent studies have successfully integrated temporal and spatial features to improve recognition accuracy. This study presents a novel approach that integrates parallel convolutional neural networks (CNNs) with a Transformer encoder and incorporates a collaborative attention mechanism (co-attention) to extract spatiotemporal features from audio samples. The proposed model is evaluated on multiple datasets and uses various fusion methods. The parallel CNNs combined with a transformer and hierarchical co-attention yield the most promising performance. In version v1 of the ASVP-ESD dataset, the proposed model achieves a weighted accuracy (WA) of 70% and an unweighted accuracy (UW) of 67%. In version 2 of the ASVP-ESD dataset, the model achieves a WA of 52% and a UW of 45%. Furthermore, the model was evaluated on the ShEMO data set to confirm its robustness and effectiveness in diverse datasets, achieving a UW of 68%. These comprehensive evaluations across multiple datasets highlight the generalizability of the proposed approach.
ISSN:2376-5992
DOI:10.7717/peerj-cs.3254
Fuente:Advanced Technologies & Aerospace Database