Deep Neural Network Regression Based Device Free Localization Technique in Changing Indoor Environment

Conventional device free localization techniques are more of a classification approach in which the target location is identified as a probable region. For small room setups and datasets, classification-based device free localization can be used conveniently to find the target’s location area. However, many application scenarios demand device free localization algorithm that can accurately find the location coordinates of the target rather than the probable region. Also, finding the exact location is another tedious task for large spaces and limited datasets. For such scenarios, regression based device free localization is an obvious choice. Deep learning in device free localization is becoming popular due to its potential merits in finding the location coordinates accurately in real time scenarios. The proposed link distance based device free localization using a deep neural network is found to be low complex and highly accurate. The highest mean localization error is 21cm, with a mean execution time of 0.0567 ms.