Order and phase ambiguities correction in the ICA based separation of speech signals

Independent component analysis (ICA) blindly separates mixed recorded signals and is employed in various engineering disciplines like speech, biomedical, communication, robotics, leakage detection, vibration analysis and machinery fault diagnosis. Order and phase indeterminacies exist in the ICA post processed signals. These indeterminacies appear due to the use of non-linearities in the development of the ICA algorithms which also restrict the practical applications of ICA. This issue of indeterminacies is not very well explored in the research. In the literature, the signals order ambiguity is resolved using various criterion that can order the signals in ascending or descending order based on a defined criteria. Although, in practice, it is required that the separated signals should be in the order of the mixing source signals. Moreover, the phase correction performed in the literature is through complicated signal processing techniques. In this paper, we propose an efficient and more rational signals ordering and phase correction technique. The proposed technique orders the separated signals in the order of the mixing source signals that makes the proposed technique more practical. Further, we have analysed the resultant mixing and un-mixing matrices, to overcome the phase and order ambiguities in the ICA post processed signals. Audio signals have been utilized for simulation purpose. Furthermore, we have utilize the FastICA algorithm to observe the effectiveness of the proposed signals ordering and phase correction technique (SOPCT). The simple mathematical formulation and efficient simulation performance make the SOPCT technique, a best choice for the practical scenarios.