2024/05/30 by Tianyu Song, Linh Thi Hoai Nguyen, Tôn Việt Tạ +1
Computer Science · Psychology · #Music and Audio Processing #Phonetics and Phonology Research #Speech Recognition and Synthesis
paper · doi:10.1016/j.csl.2024.101676
openalex publication_date 2024/05/30 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
This paper presents three innovative deep learning models for English accent classification: Multi-task Pyramid Split Attention- Densely Convolutional Networks (MPSA-DenseNet), Pyramid Split Attention- Densely Convolutional Networks (PSA-DenseNet), and Multi-task- Densely Convolutional Networks (Multi-DenseNet), that combine multi-task learning and/or the PSA module attention mechanism with DenseNet. We applied these models to data collected from five dialects of English across native English-speaking regions (England, the United States) and nonnative English-speaking regions (Hong Kong, Germany, India). Our experimental results show a significant improvement in classification accuracy, particularly with MPSA-DenseNet, which outperforms all other models, including Densely Convolutional Networks (DenseNet) and Efficient Pyramid Squeeze Attention (EPSA) models previously used for accent identification. Our findings indicate that MPSA-DenseNet is a highly promising model for accurately identifying English accents.