ARCHIVED - Helping Investigators Hear More Clearly
Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please contact us to request a format other than those available.
March 04, 2004— Ottawa, Ontario
NRC researchers are in the midst of a project that some day could have a significant impact on air safety. Air accidents, while rare in comparison to the number of flights in the air each day, do occur. According to Transport Canada statistics, the five-year average for accidents involving Canadian-registered aircraft is 339.6 accidents per year, or 9 accidents for every 100,000 hours of flying time. These accidents force us to confront and examine new information about something previously thought safe, especially in cases where there is no clear and identifiable cause.
|Cockpit, Convair 580.|
Accident investigators have a number of tools at their disposal, such as the well-known "black box", which records literally hundreds of different pieces of information about the aircraft while in flight, information which is used to help reconstruct events leading up to an accident. All planes also have a series of microphones in the cockpit to record conversations and commands from the air crew. Three audio channels are reserved for pilots, while the fourth is used for what is known as the Cockpit Area Microphone (CAM), which feeds into the Cockpit Voice Recorder (CVR).
The CAM is like a listening post into the activities and noises heard in the cockpit, a proverbial fly on the wall. Given this position, the CAM should be a valuable source of data for investigators. But, according to Elias Politis of the NRC Institute for Aerospace Research (NRC-IAR), this is not the case. In fact, the audio quality of cockpit recordings is less than optimal. "In a natural environment, it is very easy to listen and understand what is being said, but a cockpit environment is very different," Politis explained, highlighting the limitations of simply installing a microphone and hoping it will record something intelligible and useful to work with in the event of an accident. The CAM is fed by a single channel of audio and, due to flight engineering restrictions, is placed wherever there is room and it will not interfere with pilot operations. Sound quality is affected by a number of variables unique to aircraft, such as altitude, pressure and temperature changes, not to mention engine and wind noise.
To help establish whether the information they collect and analyze using new technology is measurably more intelligible than regular CVR data, the team is working with a tool used by acoustics researchers, known as the speech transmission index (STI). The STI is a measure of intelligibility, with 0 being completely unintelligible and 1 being perfectly intelligible. The STI uses a special test signal with speech-like characteristics. All CVR data collected in tests using NRC-IAR's Convair 580 will contain the STI signal, which will give the team an accurate and unbiased way of measuring if their experiments result in increased intelligibility.
Politis, who runs the NRC-IAR Flight Recorder Playback Centre, hopes to improve the audio quality of CVR data and, in doing so, create useful data for investigators to use. Politis points out that the human ear will always provide the best information because its ability to focus on key events. "The human ear is a very effective filter in that it picks up what it wants to hear. But, a microphone is very indiscriminate and simply picks everything up," he said. In the case of an accident investigation, all of this "picked up" data can be a hindrance because it has the potential of masking important data.
Politis' research project is applying several methods to deal with these factors and improve the quality of data captured by the CAM. To begin, the team is experimenting with a number of microphone types (multi-distributed microphones and microphone arrays) and cockpit placement options. While there will still be restrictions on where the CAM can be placed, researchers hope to propose several placement options which, when used, demonstrate a measurable improvement in sound quality. To really improve the usefulness of CVR data, it must be handled better, in effect, by mimicking the abilities of the human ear to filter out unwanted data and focus on meaningful events. The goal is to take something which, at the present time, is simply unintelligible, and create data where analysis and interpretation is possible.
Researchers have investigated noise reduction strategies used in other applications, such as cell phones and hearing aids and are working towards refining algorithims used by these applications to reduce noise. According to Politis, investigators find their answers in data which is suddenly different from the norm, in other words, events that stand out from the background noise. The challenge is in removing information that is clearly static and unchanging and keeping a record of important transient events. By isolating this type of information from the wall of sound found in the cockpit, the team feels this material will become more intelligible. But, said Poitis, "You have to demonstrate to the regulators that you are not taking away too much."
|NRC-IAR Team members. Left to right, Ashraf Othman, Elias Politis, Stephen Ortis.|
The team is currently working with a large sample of cockpit voice recordings collected as part of a regular series of flights for another NRC-IAR research project involving ice storms (see link at the bottom of the page for more information on this project). Politis hopes to use this information to generate several different CAM-filtering algorithms optimized to different cockpit configurations. The overall goal is to help increase air safety. "Our ultimate aim is to give investigators a better tool. Statistically, there are actually not that many cases where the CVR is a significant factor. But, there are cases where, clearly, improved data from the CVR would have played an important role. There only has to be one of these cases and we have our payback."
Enquiries: Media relations
National Research Council of Canada
Report a problem or mistake on this page
- Date modified: