An Audio Processing Approach using Ensemble Learning for Speech-Emotion Recognition for Children with ASD
Bibliographic record
Abstract
Children with Autism Spectrum Disorder (ASD) find it difficult to detect human emotions in social interactions. A speech emotion recognition system was developed in this work, which aims to help these children to better identify the emotions of their communication partner. The system was developed using machine learning and deep learning techniques. Through the use of ensemble learning, multiple machine learning algorithms were joined to provide a final prediction on the recorded input utterances. The ensemble of models includes a Support Vector Machine (SVM), a Multi-Layer Perceptron (MLP), and a Recurrent Neural Network (RNN). All three models were trained on the Ryerson Audio-Visual Database of Emotional Speech and Songs (RAVDESS), the Toronto Emotional Speech Set (TESS), and the Crowd-sourced Emotional Multimodal Actors Dataset (CREMA-D). A fourth dataset was used, which was created by adding background noise to the clean speech files from the datasets previously mentioned. The paper describes the audio processing of the samples, the techniques used to include the background noise, and the feature extraction coefficients considered for the development and training of the models. This study presents the performance evaluation of the individual models to each of the datasets, inclusion of the background noises, and the combination of using all of the samples in all three datasets. The evaluation was made to select optimal hyperparameters configuration of the models to evaluate the performance of the ensemble learning approach through majority voting. The overall performance of the ensemble learning reached a peak accuracy of 66.5%, reaching a higher performance emotion classification accuracy than the MLP model which reached 65.7%.
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 | 0.001 | 0.002 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
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".