FedEmoNet: Partitioned Data for Federated Speech Emotion Recognition Experiments

5.5 KB of partitioned data supporting a federated learning framework for speech emotion recognition. The dataset, authored by Mohammed Tawfik and last updated in May 2026, is associated with experiments on three corpora: EmoDB (German), RAVDESS (English), and CREMA-D. It likely contains client assignments and experimental splits for non-IID federated training and evaluation.

Use Cases

• Benchmarking federated learning algorithms for speech emotion recognition based on the described non-IID client partitioning.
• Evaluating cross-corpus generalization performance based on the described zero-shot transfer to CREMA-D.
• Analyzing the impact of differential privacy on model utility and membership inference attack resistance based on the described experimental results.
• Studying feature importance for emotion recognition across languages based on the described SHAP-LIME explainability analysis.

Strengths

• Includes formal differential privacy guarantees, with membership inference attack resistance dropping to an AUC of 0.52.
• Supports cross-corpus evaluation, reporting zero-shot accuracy of 68.15% on CREMA-D.
• Framework achieves high within-corpus accuracy, reported as 99.07% on EmoDB and 98.96% on RAVDESS.

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 at 5.5 KB, indicating it contains partitioning metadata rather than raw audio or feature data.

Provenance

Source

figshare

Collection Method

Likely generated as part of the FedEmoNet research framework for experimental evaluation.

Time Range

null

Freshness

Last updated 2026-05 17:45:41; freshness should be verified.

Geography

null