| Summary: | In this work, two interpretation models of Colombian Sign Language (CSL) are presented, using static and dynamic methods that employ artificial intelligence. The CRISP-DM methodology was used as a reference, creating a database with videos from seventy non-expert participants, being preprocessed and subsequently divided into proportions of 70% - 30% for training and testing, respectively. The repository was named LSC-W70 and was used on a pre-trained model of convolutional neural networks and another in combination with LSTM networks. The results reached 67% and 76% accuracy for the static and dynamic models, respectively, where the dynamic model presents improvements in similar signs by identifying the direction of movement to define the type of sign. In this sense, a dynamic Colombian sign language interpretation tool was developed that helps close communication gaps, generating equality between people.
|
|---|
