SIGNO-PERU-ISO: A Minimal-Pair Dataset of Isolated Signs for Peruvian Sign Language Recognition

The SIGNO-PERU-ISO dataset is a comprehensive multimodal collection of 18,179 RGB-D video sequences designed for the recognition of Peruvian Sign Language (LSP) across six semantic categories. Developed using an Azure Kinect DK device, the dataset captures the performances of 17 native signers in a controlled environment, preserving natural variations in speed and execution style. Its technical strength lies in the integration of three data streams: high-definition RGB video ($1920 \times 1080$) for handshapes and facial expressions, depth maps ($512 \times 512$) to resolve spatial ambiguities and occlusions, and skeletal tracking of 32 body joints. This multimodal approach provides a robust framework for analyzing both manual and non-manual components of LSP through high-fidelity visual and volumetric data. The dataset is only used for research. The data includes identifiable facial information of LSP signers. Therefore, it is not made available to the public. However, researcher(s) interested in using the SIGNO-PERU-ISO dataset for research purposes can make a request by completing the Data Usage Agreement [DUA] provided in the URL link below: https://drive.google.com/file/d/1fxKhho_wPVPu48rsdDWN5vfqOsy0r8ZF/view?usp=sharing