In the field of array signal processing, direction of arrival (DOA) estimation is a hotspot problem. Classical DOA estimation methods usually require the number of sensor should be larger than the source signals’ (which the so-called over-determined case is). However, what we encounter in practice is always the underdetermined case in which the number of source signal is larger than the sensors’. To solve the problem, a multiple signal classification (MUSIC) extension algorithm based on spatial time-frequency distribution was proposed to achieve the underdetermined DOA estimation by expanding the dimension of the spatial time-frequency distributions matrices. Compared with the existing time-frequency MUSIC, the proposed algorithm can be applied to both the over-determined and the underdetermined cases. The proposed algorithm also has advantages over the existing underdetermined DOA estimation methods for it guarantees the estimation precision, relaxes the requirements for source signal sparseness and lowers standards of the number of snapshots. Simulation results confirm the validity and high performance of the proposed algorithm.