Speech Intelligibility in Cockpit Voice Recorders

摘要

This paper focuses on the issue of correcting failures in analog open reel tape Cockpit Voice Recorders when dealing with speech intelligibility. The failure of the erase function in Cockpit Voice Recorders is a common problem. This issue causes speech to be recorded on top of a prior recording without erasure, thereby rendering either recording unintelligible. The current study investigates the potential benefit of models for speech intelligibility in providing guidance for use of cockpit voice recorders. The use of computational speech intelligibility systems could remedy this problem. This paper focuses on the Articulation Index, the Speech Intelligibility Index, and the Speech Transmission Index. These all use the Signal-to-Noise Ratio to help measure intelligibility. This paper determines a threshold for the Signal-to-Noise Ratio and intelligibility with layered continuous human speech to improve functionality of cockpit recordings.