Energy detection is widely used by cognitive radios for spectrum sensing. During a silent period, secondary users (SUs) are kept silent so that the energy detector does not confuse SU signals for primary user (PU) signals. Due to imperfect coordination, an SU may transmit during a silent period and cause possible false alarms. We propose to leverage matched filters that already exist in many SUs to alleviate the impact of such SU interference by combining the matched filtering result and the energy detection result. The analysis shows that for practical purposes, our algorithm virtually eliminates all of the negative impact of SU interference with only negligible penalty in delay and energy consumption.