Abstract: A novel class of bounded statistics, namely, the M-estimate like correlation (MELC) for independently identical distributed symmetric alpha-stable ( ) random variables was defined. Based on the MELC matrix for the array sensor outputs, a new algorithm for direction of arrival (DOA) estimation in the presence of complex noise was proposed. The comprehensive Monte-Carlo simulation results show that the MELC-MUSIC algorithm not only outperforms the fractional lower order statistics (FLOS) based MUSIC algorithms under low SNR conditions and multi-source signals environment, but also is robust with circular and noncircular signals.