A combined blind source separation and adaptive noise cancellation scheme with potential application in blind acoustic parameter extraction

Room acoustic parameters such as reverberation time (RT) can be extracted from passively received speech signals by certain ‘blind’ methods, thereby mitigating the need for good controlled excitation signals or prior information of the room geometry. Observation noise will, however, degrade such methods greatly. In this paper we therefore propose a new method, which utilizes blind source separation (BSS) and adaptive noise cancellation (ANC) to remove the unknown noise from the passively received reverberant speech signal, so that more accurate room acoustic parameters can be extracted from the output of the ANC. As a demonstration we utilize this method in combination with a maximum-likelihood estimation (MLE) based method to estimate the RT of a synthetic noise room. Simulation results show that the proposed new approach can improve the accuracy of the RT estimation in a simulated high noise environment. The potential application of the proposed approach for realistic acoustic environments is also discussed, which motivates the need for further development of more sophisticated frequency domain BSS algorithms.