Towards Automated Interpretation of Selected Nigerian Sign Language Using Multiclass Classification

Nigerian Sign Language (NSL) is the mode of communication among 16 million persons living with hearing impairment. One important development is that these sign languages differ slightly, meaning that sign languages differ from country to country. The implication is that a technology developed to help in a country for those with hearing impairment might not work out in another. In this work, we created a dataset for 21 words in the Nigerian Sign Language. We then trained a computer vision model using a multiclass classification deep learning model to identify words that could be used to form sentences. When a hearing-impaired person stands before a computer, our developed model would be able to classify each sign language, convert each sign to words, and then form a sentence from a collection of sign languages. The analysis of the proposed computer vision model shows promising results.