This work deals with problem of improving the peak-to-mean average power ratio (PAPR) of the orthogonal frequency division multiplexing (OFDM) signals. Clipping is simple and efficient way to overcome the large PAPR in OFDM systems ; however this technique results in bit error rate (BER) increasing in the receiver. In this paper we propose improvement of the iterative algorithm that is usual approach for PAPR reduction. Proposed scheme is based on soft estimation of the samples that are estimated as clipped. The performance of the proposed technique is evaluated for additive white Gaussian noise (AWGN). Simulation results shows improvement in comparison with known method without increasing complexity.