In order to reduce the pilot number and improve the spectral efficiency, recently emerged compressed sensing (CS) technique is applied for digital broadcast channel estimation. According to the six channel profiles of the ETSI digital radio mondiale (DRM) standard, the subspace pursuit (SP) algorithm is employed for the delay spread and attenuation estimation of each path in the case where the channel profile is identified and the multipath number is known beforehand. The stop condition for SP is that the estimated sparsity equals the multipath number. For the case where the multipath number is unknown, the orthogonal matching pursuit (OMP) algorithm is employed for channel estimation, while the stop condition is that the estimation satisfies the level of the noise variance. Simulation results show that with the same number of pilots, CS algorithms with randomly placed pilots outperform traditional cubic-splineinterpolation-based least square (LS) channel estimation. SP is also demonstrated to be better than OMP when the multipath number is known as a priori.