FedEmoNet: Computational Efficiency Metrics for Federated Speech Emotion Recognition

5.5 KB of computational efficiency metrics from the FedEmoNet framework, authored by Mohammed Tawfik and last updated in May 2026. The dataset likely contains performance metrics from a federated learning system for speech emotion recognition, evaluated on German (EmoDB) and English (RAVDESS) speech corpora. The framework achieved high accuracy on held-out test sets and was tested for cross-corpus generalization on CREMA-D.

Use Cases

• Benchmarking federated learning efficiency based on reported accuracy and privacy metrics.
• Analyzing cross-corpus generalization performance based on the reported zero-shot evaluation results.
• Studying the impact of differential privacy on model utility and attack resistance based on the described membership inference results.
• Investigating feature importance for emotion recognition based on the described SHAP–LIME explainability analysis.

Strengths

• Includes specific performance metrics: 99.07% accuracy on EmoDB and 98.96% on RAVDESS.
• Evaluates cross-corpus generalization with a reported 68.15% accuracy on CREMA-D.
• Provides results for differential privacy impact, showing membership inference attack resistance (AUC = 0.52) while retaining 98.5% accuracy.

Limitations

• Row count is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small (5.5 KB), indicating a limited scope of metrics rather than raw data.

Provenance

Source

Mohammed Tawfik via figshare.

Collection Method

Likely extracted from experimental results of the FedEmoNet framework combining FedProx, TCN-Transformer, PSO, and differential privacy.

Time Range

null

Freshness

Last updated 2026-05-07 17:45:50; freshness should be verified.

Geography

null