A system (200) and method (400) for near-end detection of voice (107) in speakerphone mode is provided. The method can include determining (402) a convergence of an adaptive filter (220), determining (404) a dissimilarity between an autocorrelation (311) of an echo estimate (244) and an autocorrelation (312) of a microphone signal (243) if the adaptive filter has converged, computing (406) a weighting factor (279) based on the dissimilarity, applying the weighting factor to a voice activity level (281) to produce a weighted voice activity level (283), comparing (410) the weighted voice activity level to a constant threshold, and performing (412) a muting operation in accordance with the comparing for providing half-duplex communication.