Automatic Sign Language Recognition (ASLR) is an area of active current research that aims to facilitate communication between deaf and hearing people. Recognizing sign language, particularly in the context of Algerian Sign Language (ALGSL), presents unique challenges that have yet to be comprehensively explored. So far, to the best of our knowledge, no study has considered the ALGSL Recognition. This is mainly due to the lack of available datasets. To overcome this challenge, we propose the ALGSL89 dataset, a pioneering effort in ALGSL research. The ALGSL89 dataset encompasses 4885 videos, capturing 89 distinct ALGSL signs, recorded by 10 subjects. This dataset serves as a foundational resource for advancing ASLR research specific to the Algerian signing community. In addition, we provide a comprehensive analysis of its characteristics, including statistical insights and detailed information on handshapes, positions, trajectories, and the dynamic aspects of sign movements. These details are crucial for researchers to gain a nuanced understanding of the dataset, ensuring its effective utilization in ASLR studies. In order to test the validity of our dataset, we provide the results obtained by applying a set of deep learning models. Finally, we present SignAtlas, an innovative ALGSL recognition system based on Autoencoder model.