Deep Learning based Human Activity Recognition in Smart Home

Human activity recognition (HAR) in the context of smart homes has attracted considerable interest because to its potential to increase residents' quality of life, safety, and energy efficiency. This study dives deeper into the use of deep learning (DL) algorithms for human activity recognition in a smart home scenario. A realistic approach for distinguishing distinct human behaviors like as walking, sitting, cooking, and sleeping is provided by the combination of artificial intelligence and sensor technology. This study offers FCN-LSTM, a revolutionary technique for HAR in smart homes that combines a Fully Convolutional Network (FCN) architecture with a Long Short-Term Memory (LSTM). The proposed FCN-LSTM model capitalises on the strengths of both architectures, combining the FCN's ability to capture specific information spatially with the LSTM's experience modelling temporal correlations. This hybrid method efficiently captures both spatial and temporal variability inherent in human activities, overcoming the HAR issues. The FCN-LSTM model is trained on a huge dataset acquired from sensors deployed in a smart home setting, which includes a wide range of actions encompassing ordinary daily routines. This method is used to the Wireless Sensor Data Mining (WISDM) dataset, which was produced from data collected from a large number of people who engaged in six different activities: walking, sitting, downstairs, running, standing, and upstairs. According on the experimental data, the FCN-LSTM model outperforms established approaches and standalone systems. The results show that the FCN-LSTM algorithm performs well in recognizing human actions. The FCN-LSTM algorithm achieves a testing precision of 95.99%, with recall and f-measure of 97.11% and 99.99%, respectively.