In this paper a scheme is presented for enhancing speech that is severely interfered with by undesired periodic interference. The proposed scheme structure depends on the estimation of the interference using an adaptive line enhancer whose output represents a reference input to an adaptive noise canceller. The output of the noise canceller represents a course estimate of the desired speech signal with a moderate output signal to noise ratio. Then the output signal of the adaptive noise canceller is applied to another adaptive line enhancer to obtain a good estimate of the speech signal with high improvement factor in the output SNR of more than 40 dB.