Miners' Unsafe Emotions Database for Bimodal Recognition

A dataset containing five categories of miners' unsafe emotions, used to train a deep learning bimodal framework. The dataset was used to achieve a mean recognition accuracy of 85.56% in a model integrating speech and facial expression features. It was created by Ying Lu and last updated in May 2026.

Use Cases

• Training emotion recognition models based on labeled categories of unsafe emotions.
• Developing safety monitoring systems based on speech and facial expression features.
• Evaluating the performance of feature enhancement strategies for model generalization.
• Conducting preliminary field tests of AI models in real-world industrial environments.

Strengths

• Dataset is used to achieve a reported mean recognition accuracy of 85.56%.
• Includes data from a preliminary field test in a real mining environment.
• Released under a permissive CC-BY-4.0 license for reuse.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

Ying Lu via figshare.

Collection Method

Constructed for network training of a proposed deep learning framework.

Freshness

Last updated 2026-05-15 17:36:58; freshness should be verified.