Power line communication (PLC) channel as a transmitting medium for high speed data communication based on orthogonal frequency division multiplexing (OFDM) is demonstrated. High peak to average power ratio (PAPR) in OFDM, frequency selectivity and impulsive noise in PLC channel are the most important challenges that should be taken into consideration. In this paper, a new SFTS algorithm with low searching complexity is applied to popular PTS method to reduce the PAPR parameter in which there is no need to send any side information. Also, the robust relevance vector machine based channel estimation is jointly used with the PAPR reduction. Our proposed channel estimation algorithm is based on sparse Bayesian learning with a sigmoid function as a new multi-kernel with proper initial hyper-parameters. It improves BER and MSE performances in PLC channels contaminated by impulsive noises. Proposed jointly PAPR reduction and channel estimation without any side information achieves good results comparing to parallel Tabu search PAPR reduction and Huang channel estimation algorithms. It is shown that in our proposed PAPR reduction algorithm about 1.5 and 4.5 dB improvements are obtained with considerable searching complexity reduction comparing to the parallel Tabu search method and conventional OFDM, respectively. Also, about 2 dB improvement achieved in SNIR for BER = 10−3 of proposed channel estimation algorithm comparing to LS method and 5 dB enhancement in MSE comparing to Huang method.