Research On FM-based Device-Free Wireless Localization And Activity Recognition System

Device-free wireless localization and activity recognition is an emerging techniquewhich could estimate the activity and the location of a person without equipping him/her with any device.It has a broad application prospects in the pervasive computing,smart home,security,rescue,and many other fields,and of great significance for improving people's living standards and protecting social security.With the widely utilization of wireless networks,people are "surrounded" by a variety of wireless signals all the time.Device-free wireless localization and activity recognition system can estimate people's activity by analyzing his/her impact on the surrounding wireless signals.Therefore,how to effectively describe his/her influence on the surrounding wireless signals is the key issue in our research.Based on this,we mainly explore this problem the following three aspects.1.We investigate and research the possibility and methods to realize device-free wireless localization and activity recognition by FM signal.In order to improve the universality of the proposed DFLAR system,we select the pervasive used FM signal.We select the conventional FM signal as our observation signal from a wide range of wireless signals to improve the universality of the whole system.During the raw data collection process,we do not have to use a special transmitter,only use an FM signal receiver to receive the signal in free space,which greatly reducing the complexity of the hardware device.2.During the process of extracting features,we construct multi-domain features.We creatively apply the spatial diversity and frequency diversity to effectively describe people's impact on the surrounding wireless signals and reduce the incidence selective fading in the meanwhile.Besides,the multi-domain features such as the time domain,frequency domain and the wavelet domain can increase the diversity and validity of the feature effectively.3.In the classification and recognition part,we select the robust sparse representation classification algorithm.We use this robust algorithm for classification for the fact that the application environment changes and other factors will introduce noise.Besides,we introduce iterative sparse representation classification algorithm to improve the classification accuracy,which makes this system a more robust system in the whole.In order to verify the performance of our FM-based device-free wireless localization and activity recognition system,we have carried out verification experiments in two typical laboratory environments.The experimental results show that the system behaves better performances in localization and activity recognition,recognition accuracy can reach more than 90% in noisy laboratory environments.