Enhancing Speech Emotion Recognition through Domain-Aware Data Augmentation and Model Explainability
Bibliographic record
Abstract
Speech Emotion Recognition (SER) constitutes a vital technology for enhancing the naturalness and intelligence of human-computer interactions. However, its performance is often constrained by noise interference and data scarcity in real-world contexts. To address these challenges, this study systematically investigates the construction of high-precision and interpretable SER models utilizing the Toronto Emotional Speech Set (TESS) dataset. A data augmentation strategy, rooted in domain knowledge, is developed, encompassing dynamic signal-to-noise ratio adjustment, emotion label-guided speech rate modification, and pitch transformation, to produce emotionally congruent augmented samples. Subsequently, a comprehensive 134dimensional acoustic feature vector is extracted, incorporating Mel Frequency Cepstral Coefficients (MFCCs), Mel spectrogram, spectral contrast, and chroma features. The classification performance of six models—Logistic Regression, Decision Tree, Random Forest, Extreme Gradient Boosting, Long Short-Term Memory, and Transformer—is rigorously evaluated based on these features. Empirical findings indicate that the Extreme Gradient Boosting model achieves superior performance, with both accuracy and macro F1 score surpassing 99.4%. Further Shapley Additive Explanation identifies the Mel spectral energy distribution as the most critical discriminative feature for the model, followed by the MFCC, aligning with the physiological acoustic mechanisms of speech production. This study not only highlights the exceptional potential of lightweight machine learning models in this domain but also offers an empirical foundation and critical guidance for developing reliable and transparent SER systems through its comprehensive framework of data augmentation, feature engineering, model selection, and decision interpretation.
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 | 0.002 | 0.005 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.002 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
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".