Recently, use of automatic speech recognition system is demanded for various applications such as security, word to text conversion etc. During the speech signal acquisition, other unwanted signals from various sources are added to the original signal which degrades the performance of ASR system. These unwanted signals are called as noise or mixing of sources which are caused due to multi-user recording, echo effect etc. this issues motivates to develop an efficient algorithm for audio demixing or source separation. To address this issue in this work we propose a new approach for source separation method using nonnegative factorization method. Proposed work utilized source mixing signal modelling, filter bank designing and source separation algorithm implementation. Modelling of signal is performed by combining two different channels which are acquired from different source, this signal is called mixture signal. Later a filter bank is designed using scattering algorithm based on wavelet transform method and a optimization problem is formulated for audio demixing. Experimental study shows the robustness of proposed model by considering various implementation scenarios.