Mexican Sign Language Corpus with 121 Dynamic Gestures for Medical Emergencies

A corpus of 121 dynamic signs from Mexican Sign Language (LSM) was developed in collaboration with expert interpreters, with a specialized lexicon for medical emergencies and accident scenarios. The dataset includes videos recorded using a standardized protocol with expert and non-expert participants. The work by Jesús Antonio Navarrete-López, last updated in April 2026, evaluated multiple deep learning architectures for gesture recognition.

Use Cases

• Training deep learning models for dynamic sign language recognition based on skeletal keypoint data.
• Developing assistive technologies for LSM-to-Spanish translation based on the specialized medical emergency lexicon.
• Benchmarking performance of architectures like ResNet, LSTM, and Transformer models on video-based gesture data.
• Studying the impact of data augmentation and normalization techniques on sign recognition model generalization.

Strengths

• Includes 121 distinct signs with a specialized focus on medical emergencies and accidents.
• Uses a standardized video acquisition protocol with both expert and non-expert participants.
• Model evaluation reports a high F1-score of 0.948 for the best-performing ResNet architecture.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is described in a 141.6 KB PDF, indicating the primary data files are not directly available or the scope is limited to the paper's results.

Provenance

Source

figshare, author Jesús Antonio Navarrete-López.

Collection Method

Videos recorded using a standardized protocol in collaboration with expert interpreters; skeletal keypoints extracted using MediaPipe.

Time Range

null

Freshness

Last updated 2026-04-16 05:13:53; freshness should be verified.

Geography

Mexico (Mexican Sign Language).