Abstract: A smoothed L0 (SL0) algorithm based on compressed sensing proposed in previous works for inter symbol interference (ISI) sparse channel estimation. But this method has “notched effect” due to the negative iterative gradient direction. Moreover, the “steep nature” of cost function in SL0 is not steep enough, leading to channel estimation errors and make convergence results not the most optimal. The lagrange multipliers and newton method were combined to optimize SL0 algorithm in order to obtain a more rapid and efficient signal reconstruction algorithm termed as an improved smoothed L0 (ISL0). The channel state information (CSI) of the sparse multi-path channel was obtained and analysis of reconstructed signal deviation, mean squared error (MSE) in the perspective of iterations and signal-to-noise ratio (SNR) as well as the iteration time and ISI equalization performance were also done. Furthermore, the superiority of ISL0 has been verified by computer simulation. Real-time simulation results clearly show that the ISL0 algorithm can estimate the ISI sparse channel much better. Compared with CoSaMP, SL0 and some other algorithms, the ISL0 algorithm can greatly improve the performance of system in the same channel environments.