Artificial intelligence based real time human sign morphological recognition in smart homes

Real-time detection of sign language is one of the solutions that can greatly help in changing the quality of life of individuals or persons with communication or physical disabilities in the rapidly changing world of smart homes. The paper represents a new proposal for using Long Short-Term Memory (LSTM) networks in identification of human activities and sign language. A variation of recurrent neural networks called LSTM is very suitable to be applied to sequential data, so it would be the right candidate to process time-series data produced by sensor-enabled smart homes. With the help of the time connections that LSTM records, the proposed system can effectively detect daily activities and interpret sign language gestures with a high degree of precision based on the ResNet structure. The technology of an integrated system in smart home placements can hopefully result in higher automation, safety, and accessibility systems offering a more natural relationship between individuals and their surroundings, with accuracy reaching 94.67%. Experimental results show that LSTM model accuracy rates in both human activity and sign language recognition and an inference can be made that the LSTM model can be used in large-scale systems in smart homes.