Optimising Wi-Fi Detection Systems: A Novel Approach to Subcarrier Selection in Dynamic Environments

People at risk of falling can rely on in-home health care solutions to detect incidents like falls and ensure assistance is promptly summoned. Current systems make use of wearable devices, which potentially may not be always worn due to comfort, battery life or forgetfulness. Another solution is contactless methods such as Wi-Fi sensing being used to detect movements. This paper looks at one of the challenges of Wi-Fi sensing, which is the changing of locations. Wi-Fi sensing makes use of Artificial Intelligence (AI) algorithms to recognise changes in the signal propagation associated with a specific movement, such as a fall. However, when changing the location, the signal propagation changes and thus impacts on any previously trained algorithm. This paper proposes a method of subcarrier selection, which is consistent in identifying sitting and standing movements within multiple locations. Without this method, the best accuracy from the selected algorithms across four different locations was 66.15 %. Results improved to 93.49 % when using the proposed method. These results show how important novel signal processing techniques are in improving the accuracy of important contactless healthcare technologies.