The dynamic Colombian sign language dataset for basic conversation LSC70

Abstract

Sign language is a form of non-verbal communication used by people with hearing disability. This form of communication relies on the use of signs, gestures, facial expressions, and more. Considering that in Colombia, the population with hearing impairments is around half a million, a database of dynamic, alphanumeric signs and commonly used words was created to establish a basic conversation. For this purpose, 70 non-expert volunteers in Colombian Sign Language participated, and they were recorded against a clear background under uncontrolled lighting and clothing conditions. The dataset was named LSC70 and includes a six-frame transition in JPG format. It is organized into three parts: The first part contains alphanumeric signs at a resolution of 640×480 pixels; the second part includes word signs at a size of 640×480 pixels; and the third part focuses on the dominant hand in alphanumeric signs at 120×120 pixels. Through filtering that removed similar signs, the dataset now contains 35,208 frames with a total of 47 different signs. These images have been validated by experts at the University of Cauca and are available for free use within the research community.