SwinTSER: An Improved Bilingual Speech Emotion Recognition Using Shift Window Transformer
Bibliographic record
Abstract
Abstract Emotion recognition from human speech occupies a significant position in Human-Computer Interaction, especially with the recent advancements in Artificial Intelligence and Robotic computing. As the level of interactivity of man–machine increases, intuitive responses that are emotionally based have attracted a lot of research into emotion recognition from speech signals. However, with various machine learning models littering the literature, cross-language efficient speech emotion recognition with extracted features inherent in speech signals with state-of-the-art deep learning techniques, is still posing a serious challenge. In this paper, we proposed a deep learning transformer network based on a shift window for speech emotion recognition using speech corpus from two different languages. Shift Window Transformer (SWT) is based on a hierarchical transformer architecture designed for natural language tasks and has recently become a novel model in computer vision and image processing tasks. The input feature to the model, Mel spectrogram, is extracted from two public speech datasets: Toronto English Emotion Speech (TEES) and EMOVO. Our proposed transformer model achieved a promising result of 98.3%, 64%, and 66% recognition accuracy on TESS, EMOVO, and TESS_EMOVO (hybrid bi-lingual) datasets, respectively, after extensive experiments and parameter optimization. Our performance evaluation revealed that the proposed model yielded an improved result in the recognition of six different emotions from human auditory speech compared to others found in the literature. The study explores the performance of the SWT architecture on cross-language speech emotion recognition and informs future robust and adaptive model development.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".