The minimum-symbol-error-rate (MSER) decision feedback equalizer (DFE) has been developed for underwater channels. Comparing with conventional normalized least-mean-square (NLMS) with DFE, MSER-DFE shows significant performance improvement in terms of bit-error-rate (BER). However, it suffers from slow convergence. In this paper, to improve the convergence speed of MSER-DFE, we modify the objective function to exploit the sparsity of the channel equalizer and use the sparseness measure, which is referred to as sparse control proportionate minimum-symbol-error-rate (SC-PMSER) DFE. By assigning larger weights to equalizer taps with larger values, we can speed up the convergence of the proposed equalizer at the cost of higher computational complexity. Simulation results show the effective of the proposed detector.