With the advancements in the mobile devices having Bluetooth Low Energy (BLE) capability, the BLE based indoor target localization is the recent trend. The majority of indoor localization methods generally rely on traditional simple techniques such as trilateration or angulation. However significant localization errors are involved with these techniques due to highly nonlinear relationship between RSSI and distance because of issues such as NLOS, multipath propagation. The Generalized Regression Neural Network (GRNN) with a one pass learning capability, is well known for its ability to train quickly. This paper proposes an application of GRNN as an alternative to these traditional techniques to obtain first location estimates of moving person using a hybrid network of PSOC BLE nodes and smartphone, which are further refined using Kalman filtering (KF) framework. Two algorithms namely, GRNN + Kalman filter and GRNN + Unscented Kalman filter are proposed in this research work. The GRNN is trained with the RSSI values from PSOC BLE nodes at various locations and the corresponding actual 2-D locations of the given monitoring area. The real time experiments prove the efficacy of the proposed algorithms over the traditional approach in the context of NLOS, multipath propagation.