Research on High-Quality Data Screening Algorithm for Sign Language

This paper introduces the collection process of a Chinese sign language dataset for practical applications, and makes a statistical analysis of word frequency, sentence frequency and other data information. For this dataset, and for the subsequent larger-scale sign language data collection, an artificial intelligence application-oriented sign language video high-quality data screening algorithm is proposed to realize an intelligent solution for machine screening of sign language video quality through the process of noise filtering, multi-feature extracting, and screening framework designing. It is experimentally verified that the solution can completely screen out misplaced videos, but it is consistent with the sensitivity of human eyes to observing hand changes, it is not easy to distinguish subtle omissions, wrong sign and redundant sign, and the order exchange. This scheme can basically meet the screening requirements in the subsequent large-scale sign language video recording process, which can increase the robustness of the data set and improve the data quality while saving labor costs, thus providing a high-quality data base for establishing an accurate sign language recognition and translation model.