<p>The need for indoor localization is significantly growing, especially for locating people inside huge buildings, tracking products in large warehouses, and optimizing next-generation wireless networks by minimizing latency. The multipath phenomenon is one of the most challenging issues encountered in any indoor positioning system requiring the propagation of an electromagnetic wave in the localization area. Existing localization methods provide either low precision localization score with minimal latency or precise localization with high latency, such as the case of convolutional neural network-based methods, which require high computing power. In this paper, we propose a new localization method that overcomes the multipath phenomenon while being computationally attractive. The method is based on spectral analysis for estimating the overlap duration and using it as a feed-forward back propagation neural network feature. The overlap estimator and the localization accuracy are evaluated using simulations and real-world experiments. Results demonstrate that the proposed method provides high localization accuracy and requires low computing power compared to existing methods.</p>