Effect of Data Augmentation, Cross-Validation Methods in Robustness of Explainable Speech Based Emotion Recognition
Bibliographic record
Abstract
Robust and interpretable speech emotion recognition (SER) is pivotal to handle real time noisy conditions as well as to adapt the language and speaking style variations.In this work, the proposed feature vector is validated using the German database EMoDB for speaker dependent (SD) and speaker independent (SI) approaches.Data augmentation techniques of time shifting, pitch shifting and adding Gaussian noise are employed along with data balancing.The baseline feature set includes spectral features of mel-frequency cepstral coefficients (MFCC) and log Mel spectrogram-based features along with time domain features of zero crossing rate and root mean square energy.Proposed reduced feature set is fusion of baseline features with discrete wavelet transform (DWT) based statistical features selected through ANOVA and reduced chroma features obtained using principal component analysis (PCA).Proposed reduced feature vector is validated by support vector machine (SVM), multilayer perceptron (MLP), and Gradient Boosting (XGBoost) classification with holdout and stratified k-fold (k=5, K=10) cross-validation.In speaker-dependent approach, with 10-fold cross-validation accuracy is improved from 73.07%, 74.94%, 71.02% to 93.18%, 94.07%, 87.15% for SVM, MLP and XGBoost classifier respectively.The contribution of discriminative features of proposed reduced feature set in emotion prediction is explained with the Shapley Additive Explanations module.
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 | 0.021 | 0.035 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| 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".