Human speech undergoes much interference in a medium. These distortions in the speech signal may leads many disadvantageous in the hearing aid application, speech separation and synthesis. So an effective application can make right turn in field of speech separation. The development of CASA (Computational Auditory Scene Analysis) is trying to reduce these defects and improve the speech applications. The possibility of separating the dominant speech from a mixture and amplifying that may be used in the hearing aid applications. In this paper, we are introducing cochlear design of filters as the channels. Segregation and the grouping are the main methods implemented in this paper. Pitch determination is done based on the response of the cochlea model and combining them using the periodicity detection. Frequency domain analysis done based on the STFT (Short Time Fourier Transform) method. Separation of the dominant speech is done by masking. This method is computationally less complex and we can obtain the better SNR (Signal to Noise Ratio) compare to other related methods available in this literature.