Human activity recognition, a dynamic field within computer vision and machine learning, aims to accurately identify and interpret human actions or activities using video data. This paper delves deeper into the domain by concentrating on Sign Language (SL) as a unique and intricate form of human activity. Sign Language (SL) is a vital communication tool, used by individuals who are deaf or hard of hearing, facilitating their interaction and engagement with others in a visually expressive manner. A lot of research and projects in SL are being done, and it has been observed that most of them are done in American Sign Language (ASL) or Indian Sign Language (ISL), but no such progress has been carried out for the other SLs. Therefore, this paper aims to present a work done in Arabic sign language (ArSL), especially Tunisian SL. The main idea is to recognize Tunisian SL and then translate it into spoken language using artificial intelligence (AI) methods. The purpose of this work is to ease the communication between individuals who are deaf and those who are not. The proposed method uses holistic landmarks with sequence key points for Tunisian sign language and trains it with a Long Short-Term Memory (LSTM) model. The results show that the proposed model can be used for real-time SL estimation, providing a better interpretation method for the deaf community. As a result, our model achieved a validation accuracy of 98% with a cross-validation technique.