Technological projects portfolio

In our daily lives, we are surrounded by wireless networks, which allow us to communicate with ease. When it comes to Wi-Fi signals specifically, it turns out that they contain enough usable information to detect and recognize human activity in a non-intrusive way, without it being necessary to deploy an additional network of sensors.

The information on the status of the channel, which describes how the signal propagates from the emitter to the receiver, is exploited to detect changes in the wireless signal reflections due to interference of waves with a person’s body. Extracting and analyzing the characteristics of these flows of data allows for human behavior recognition.

The WIFINE project takes on this passive recognition of human activity in interior spaces using signals of opportunity of Wi-Fi systems.

WIFINE employs artificial intelligence to identify events. The approach used makes it possible to accelerate the execution of automatic learning algorithms.

However, the main advantage compared to other approaches lies in the use of a simplified neural network¹, allowing it to operate in real time. WIFINE exceeds traditional deep-learning based approaches due to the speed of execution and its millisecond training times, all while maintaining a comparable level of performance.² The effectiveness and low energy consumption make it an embeddable technology, because WIFINE can run on standard microprocessors present in devices equipped with a Wi-Fi connection.

This technology can therefore be embedded in existing Wi-Fi devices and it requires no additional hardware.

1: after pre-processing (filtering and sub-sampling), the artificial neural network of only a few hundred nodes is trained under supervision.

2: tests show that the recognition accuracy for activity monitoring (lie down, fall, walk, run, sit, stand up) is around 96%. According to the state of the art established in 2021, this is the best result combining “accuracy / execution speed / training speed”.

The technology can be applied to detection of abnormal events for the assistance of elderly people, or to the analysis of behavioral habits for a family living in a smart home (home automation systems). Medical applications are also possible. For example, it could be used to monitor the respiration rate in order to evaluate sleep quality.

Finally, in the field of monitoring / security, Wi-Fi sensing could also be used to detect attacks or suspicious or violent behavior.

Additionally, Wi-Fi sensing could be used to count people in a public transport context.