On the basis of zero-crossings detection, a new method of speech segregation is proposed using these clues of Interaural time difference(ITD), zero-crossings peak amplitude(ZCPA) and zero-crossings power. The estimation of ITD is utilizing the statistical properties of zero-crossings detected from binaural filter bank outputs in order to get more reliable ITD estimation in noisy environment. Signal phase ambiguity in the high frequencies is solved by correcting the signals in the high frequency channels by means of the ITD values calculated by ZCPA algorithm in the low channels. Spline interpolation is used for sound segregation in order to make sound signal segregated correspond to the original sound signals as much as possible. The test for sound segregation in various mixture sources is made. The results show that the method can make can segregate various mixed sound sources effectively. The advantages of the proposed method are the robustness to noise, the less computational complexity and no need to train the masks for sound segregation.