The Blind Separation of Cockpit Mixed Signals Based on Fast Independent Component Analysis

Cockpit Voice Recorder (CVR) records all the sound signals related to the accident in the cockpit before the crash. Accurate analysis of the cockpit voice signal is the basis for exploring the real reason behind the flight accident, which often starts from the separation of cockpit mixed signals. In this paper, a Fast Independent Component Analysis (FastICA) method for mixed signal separation is proposed. The basic idea of the method, three classical assumptions and the independence criteria of separation results are presented. The FastICA algorithm is analyzed theoretically, and the blind separation experiment of cockpit mixed signals is realized in this paper. The simulation experiment result shows that although FastICA algorithm is complex in theory, the realize of this algorithm is concise and clear, which can quickly and effectively separate these signals, and provide research basis for accurate analysis of the causes of aircraft accidents in the later stage.