5 Sep '22
Scientists at the Moscow Lomonosov State University (MSU) have applied neural network technology to their novel methodology of Wi-Fi scanning for human being detection. At the heart of the approach are controllable gated recurrent units. This advanced gating mechanism in recurrent neural networks is expected to show more efficacy than the existing technologies. A person’s detection and the pinpointing of his or her movements may be used in a range of applications across transportation, smart house, commercial facility security, health care and other projects.
As Wi-Fi tech develops, hotspots are becoming increasingly popular as rangefinders. There are basically two fundamental approaches to detecting movements using Wi-Fi devices. The simpler one is based on what’s known as received signal strength indicator (RSSI) and the other is based on complete channel state information (CSI) analysis.
The latter is believed to provide more accurate data than the former – but there are limitations linked to Wi-Fi hotspots to be used here. The hotspots must be multi-antenna ones and have a very special kind of control software to receive CSI values – two caveats that apparently make the approach no-go when it comes to developing an off-the-shelf solution for many purposes.
Neural outshines statistical
According to the developers, statistical algorithms are widely used to tap RSSI data for detecting the presence of a human being. There are alternatives, too, based on machine learning algorithms and neural networks.
In the most recent MSU effort, improved techniques from both categories have been used. Experiments reportedly showed that the accuracy of a gated recurrent units (GRU) based neural network, applied as a neural algorithm, was higher than that of statistical algorithms.
The team hopes what they have achieved so far will lay the foundation for the development of multipurpose methodology for using Wi-Fi scanning for human movement detection.