The position of an object is an important information which is necessary to accomplish many tasks. Humans, machines and other objects emit sound waves, which can be used to find their position. Examples of applications are finding a person shouting for help in a surveillance scenario or to localize a running car engine for driver assistance systems or self-driving vehicles. The passive source localization problem can be solved by the Time Difference of Arrival (TDOA) localization technique. The signals are captured at known sensor positions, then TDOAs of signal pairs are used to localize the source. The line-of-sight (LOS) condition requires a direct path between the source and the sensors. Many approaches are based on TDOA under LOS condition. In real world scenarios the LOS condition is often violated due to obstacles blocking the direct path. This makes localization approaches necessary which can handle the non-line-of-sight (NLOS) condition.
The Goal is to accurately localize an object by its emitted waves even if there is no line-of-sight to the sensor.
There are several publications which try to compensate the error introduced by the NLOS condition. These approaches provide a position of the source which is based on a best guess using statistics or other characteristics. Our approach is to use information of the surroundings to find the position of the source more accurately in NLOS conditions or mixed LOS / NLOS conditions. In theory this approach is capable to extract the exact position of the source.