Sound Source Separation and Identification for Aircraft Cockpit Voice Recorder

ठAnalysis of the background information recorded in aircraft Cockpit Voice Recorders (CVRs) has been proposed as a complement to the analysis of onboard Flight Data Recorders in civil aircraft investigations. This paper explores the feasibility of identifying individual sound sources using an integrated signal processing technique for signal noise cancellation and sound source separation and identification. Experimental results show that at least six different sound sources, simulated as a mixed signal, can be successfully separated and identified, using Adaptive Noise Cancellation (ANC), Blind Signal Separation (BSS) and Back Propagation Neural Network techniques, from the output of a ATR-72 CVR. It is proposed that this concept be applied, not only to aviation investigations and engine monitoring, but also to early identification of abnormal sounds from other aircraft components.