Sponsor Message:
Aviation Technical / Operations Forum
My Starred Topics | Profile | New Topic | Forum Index | Help | Search 
State Of Speech Recognition Tech In The Cockpit?  
User currently offlinecomorin From United States of America, joined May 2005, 4896 posts, RR: 16
Posted (10 months 3 weeks 20 hours ago) and read 4374 times:

We have seen a proliferation of Siri-like speech recognition software in the consumer markets, and I was wondering if the technology has become reliable enough to use in civil or military aircraft. My guess is that it is not sufficiently reliable, but wonder if there are non-critical areas where speech rec might help.

Thanks for any comments and hope this is not the wrong forum to post.

47 replies: All unread, showing first 25:
 
User currently offline26point2 From United States of America, joined Mar 2010, 820 posts, RR: 0
Reply 1, posted (10 months 3 weeks 19 hours ago) and read 4310 times:

Oh, that would be a riot! Can you imagine...?

"If you would like to put the landing gear up press 1 now."...."escuchar este mensaje en español pulse 2"


User currently offlinecomorin From United States of America, joined May 2005, 4896 posts, RR: 16
Reply 2, posted (10 months 3 weeks 18 hours ago) and read 4284 times:

Quoting 26point2 (Reply 1):

LOL!

Humor aside, it seems like a technology with potential to reduce crew workload and reduce fidgeting with knobs and dials at critical times.


User currently offlineMir From United States of America, joined Jan 2004, 21557 posts, RR: 55
Reply 3, posted (10 months 3 weeks 18 hours ago) and read 4270 times:

Quoting comorin (Reply 2):
Humor aside, it seems like a technology with potential to reduce crew workload and reduce fidgeting with knobs and dials at critical times.

Limited potential. The problem is not so much the speech recognition itself (though that would have to advance significantly beyond what we have now in order to be viable) but what that recognition is going to, and how there would be a feedback loop between the pilot and that system.

There's not that much fiddling with knobs and dials during critical times anyway.

-Mir



7 billion, one nation, imagination...it's a beautiful day
User currently offlineoly720man From United Kingdom, joined May 2004, 6697 posts, RR: 11
Reply 4, posted (10 months 3 weeks 15 hours ago) and read 4186 times:

In the sad light of AF447, "recovery" may be a useful command.... if the plane could sense where it was and put itself in the appropriate attitude and throttle setting.

Otherwise, I'm not sure what else. You could have "flaps", "wheels" "lights" or things like that, but it would add another sense of detachment from what goes on. At least if you're talking to another pilot who's in the loop you're probably reasonably sure they'll respond appropriately. Saying "flaps" and then waiting for the computer to digest this and act would be a distraction when you don't need distractions.

Added to which, would every pilot have to talk to the computer before a flight to train it? Or have a USB key with the words already filed in computerspeak?



wheat and dairy can screw up your brain
User currently offlineMir From United States of America, joined Jan 2004, 21557 posts, RR: 55
Reply 5, posted (10 months 3 weeks 10 hours ago) and read 4104 times:

Quoting oly720man (Reply 4):
In the sad light of AF447, "recovery" may be a useful command.... if the plane could sense where it was and put itself in the appropriate attitude and throttle setting.

The second part of that is the hard part. But even if such a thing were possible, why tie it to voice recognition when you could tie it to a single button on the panel? No risk of not being understood that way.

-Mir



7 billion, one nation, imagination...it's a beautiful day
User currently offlineStarlionblue From Greenland, joined Feb 2004, 17017 posts, RR: 67
Reply 6, posted (10 months 3 weeks 9 hours ago) and read 4097 times:

Quoting Mir (Reply 5):
Quoting oly720man (Reply 4):
In the sad light of AF447, "recovery" may be a useful command.... if the plane could sense where it was and put itself in the appropriate attitude and throttle setting.

The second part of that is the hard part. But even if such a thing were possible, why tie it to voice recognition when you could tie it to a single button on the panel? No risk of not being understood that way

Quite. I am reminded of the F/A-18, where if the plane goes out of control arrows on the displays tell the pilot where to position the stick. No need for voice commands.



"There are no stupid questions, but there are a lot of inquisitive idiots."
User currently offlinecomorin From United States of America, joined May 2005, 4896 posts, RR: 16
Reply 7, posted (10 months 3 weeks 9 hours ago) and read 4083 times:

Quoting Mir (Reply 3):
Limited potential. The problem is not so much the speech recognition itself (though that would have to advance significantly beyond what we have now in order to be viable) but what that recognition is going to, and how there would be a feedback loop between the pilot and that system.

Thanks for your comments - I take it you do not see a role for speech rec in the control loop even if if the hit rate became high enough.

Quoting oly720man (Reply 4):
Added to which, would every pilot have to talk to the computer before a flight to train it? Or have a USB key with the words already filed in computerspeak?

Today's command-oriented speech recognition systems don't need training, like the telephony and GPS systems that ask you to interact via voice.


User currently offlineStarlionblue From Greenland, joined Feb 2004, 17017 posts, RR: 67
Reply 8, posted (10 months 3 weeks 8 hours ago) and read 4069 times:

I do think that speech recognition will enter the cockpit at some point, but it will do so in military aircraft first. There tends to be more time critical stuff in combat.

Unless we jump directly to neural interfaces that is.



"There are no stupid questions, but there are a lot of inquisitive idiots."
User currently offlineMax Q From United States of America, joined May 2001, 4441 posts, RR: 19
Reply 9, posted (10 months 3 weeks 1 hour ago) and read 4004 times:

'i'm sorry I don't understand your answer'



'I'm sorry I don't understand your answer'



And repeat 100 times.



Not a good idea..



The best contribution to safety is a competent Pilot.
User currently offlineHAWK21M From India, joined Jan 2001, 31679 posts, RR: 56
Reply 10, posted (10 months 3 weeks 1 hour ago) and read 4002 times:

How will it handle people who speak with a different accent.
I've noticed that problem exists.



Think of the brighter side!
User currently offlinecomorin From United States of America, joined May 2005, 4896 posts, RR: 16
Reply 11, posted (10 months 2 weeks 6 days 17 hours ago) and read 3961 times:

Quoting HAWK21M (Reply 10):

Good point. While you can train your personal device, it would be a problem when you have crew rotation and diversity.

Do you have speech on your smartphone, does it work well?


User currently offlineMir From United States of America, joined Jan 2004, 21557 posts, RR: 55
Reply 12, posted (10 months 2 weeks 6 days 13 hours ago) and read 3922 times:

Quoting comorin (Reply 7):
I take it you do not see a role for speech rec in the control loop even if if the hit rate became high enough.

IMO, the time it would take to verify that the speech recognition has worked as intended is similar to or greater than the time it would take to just do whatever you wanted the speech recognition to do. And doing it yourself will provide better feedback.

-Mir



7 billion, one nation, imagination...it's a beautiful day
User currently offlineStarlionblue From Greenland, joined Feb 2004, 17017 posts, RR: 67
Reply 13, posted (10 months 2 weeks 6 days 11 hours ago) and read 3892 times:

Quoting Max Q (Reply 9):
'I'm sorry I don't understand your answer'

"I'm sorry Dave, I'm afraid I cannot do that..."



"There are no stupid questions, but there are a lot of inquisitive idiots."
User currently offlineokie From United States of America, joined Jul 2003, 2999 posts, RR: 3
Reply 14, posted (10 months 2 weeks 6 days 9 hours ago) and read 3858 times:

My main issue would be that voice recognition would pick up inadvertent conversation between CA, F/O, FA or a jumpseater that would not be an actual command.

I think I will go over to a friends pad and "Crash" tonight.

I am going to "Go Around" to some bars tonight.

Mind out of the gutter now.

Just use your imagination.

Okie


User currently offlinerwessel From United States of America, joined Jan 2007, 2342 posts, RR: 2
Reply 15, posted (10 months 2 weeks 6 days 5 hours ago) and read 3825 times:
Support Airliners.net - become a First Class Member!

While I don't think speech recognition is going to be ready for the cockpit soon (both maturity and environmental issues, like high cockpit noise, are going to be issues), but there are areas where it will be easier to apply than others. Notably things that are not an immediate safety issue if they're not done quite right. Many communication and navigation tasks fall into that category. Consider: "Raccoon: comm one, Chicago center!" or "Dreamy: navigation, show next segment!" Things where keeping the pilots eyes and hands on the primary tasks (rather than flipping switches) would be a good thing. Those could be an advantage in strong turbulence as well, when moving a lot of small controls becomes more difficult (or even touch screens, as we're starting to see now).

User currently offlineMir From United States of America, joined Jan 2004, 21557 posts, RR: 55
Reply 16, posted (10 months 2 weeks 5 days 20 hours ago) and read 3767 times:

Quoting rwessel (Reply 15):
Things where keeping the pilots eyes and hands on the primary tasks (rather than flipping switches) would be a good thing.

But the times when things that aren't immediate safety issues if they're not done right get done is a relatively low-workload period. So there's plenty of attention capability that can be used to do things.

Just because something seems cool doesn't mean it should go into a product. The auto industry has been besieged by all sorts of horribly-implemented user interface items because the marketing department wants this and that whiz-bang feature that will make people think they're driving an advanced and/or futuristic car. But this isn't the auto industry - if something is going to go into a cockpit, it should go into a cockpit because it provides more benefits than the extra complexity it's going to inevitably introduce. In the case of voice recognition, I just don't see it.

Quoting rwessel (Reply 15):
Those could be an advantage in strong turbulence as well, when moving a lot of small controls becomes more difficult (or even touch screens, as we're starting to see now).

I've never had a problem manipulating controls in turbulence. I've never tried to use a cockpit touchscreen in turbulence, but I would very much hope that the avionics designers thought about that during the design process.

-Mir



7 billion, one nation, imagination...it's a beautiful day
User currently offlineairmagnac From Germany, joined Apr 2012, 307 posts, RR: 44
Reply 17, posted (10 months 2 weeks 5 days 13 hours ago) and read 3709 times:

The benefit of speach recognition would be to gain a couple milliseconds or make handling easier in adverse conditions.
However, as Mir pointed out, the time gain would be negated by the need to wait for feedback and confirm the order, as there is an additonnal communication transfer from the pilot to the machine. And again as Mir said, handling enhancement would be marginal.
On the other hand, the system would have to work against the noise background, accents, mispronounciations, stressed pilots with distorted voices, ambiguities and parasitic speach (casual chit chat). Also, it's easy to figure out how to transmit a command when there is a button with a tag right in front of you. With speach recognition the pilot has to memorize all the commands ; not a problem in most cases, but it could be an issue in a high stress environment. So I figure the buttons would have to stay, and there is no gain in terms of cockpit layout.

I just don't see how the benefits could possibly become significant enough to justify the amount of issues and risks. But that's for cockpit interfaces.
Maybe maintenance applications could make good use of it ? The background noise factor could be even more of a problem, though.


Quoting Starlionblue (Reply 8):
I do think that speech recognition will enter the cockpit at some point, but it will do so in military aircraft first. There tends to be more time critical stuff in combat.

And as the time critical stuff can mean life or death, the risk of mis-recognition of a voice command may be acceptable on a fighter. On an airliner, I'd guess not.
And military aircraft sometimes fly in extreme conditions in terms of accelerations and attitude, impeding the movement to the switches and buttons. Not so true for an airliner



Quoting Starlionblue (Reply 8):
Unless we jump directly to neural interfaces that is.
Mitchell Gant for the win !  

The funny thing is, he has a super duper neural interface to control the advanced high performance weapons, but he still has to punch in the numbers for his nav system   
Accent doesn't seem to be a problem though...






One "oh shit" can erase a thousand "attaboys".
User currently offlineMir From United States of America, joined Jan 2004, 21557 posts, RR: 55
Reply 18, posted (10 months 2 weeks 5 days 13 hours ago) and read 3699 times:

Quoting airmagnac (Reply 17):
And as the time critical stuff can mean life or death, the risk of mis-recognition of a voice command may be acceptable on a fighter. On an airliner, I'd guess not.
And military aircraft sometimes fly in extreme conditions in terms of accelerations and attitude, impeding the movement to the switches and buttons. Not so true for an airliner

The biggest difference in my mind is the fact that fighters have the potential to be flying in situations where the pilot needs to have both hands on the controls for an extended period of time, making pushing buttons difficult. That's not true of airliners.

-Mir



7 billion, one nation, imagination...it's a beautiful day
User currently offlineStarlionblue From Greenland, joined Feb 2004, 17017 posts, RR: 67
Reply 19, posted (10 months 2 weeks 5 days 9 hours ago) and read 3674 times:

Quoting airmagnac (Reply 17):
Quoting Starlionblue (Reply 8):
Unless we jump directly to neural interfaces that is.
Mitchell Gant for the win !  

The funny thing is, he has a super duper neural interface to control the advanced high performance weapons, but he still has to punch in the numbers for his nav system   
Accent doesn't seem to be a problem though...

As I recall, he had to "think in Russian". One of the reasons they chose Gant was that he was fluent in the language.

Loved that movie when I was a kid! 



"There are no stupid questions, but there are a lot of inquisitive idiots."
User currently offlineairmagnac From Germany, joined Apr 2012, 307 posts, RR: 44
Reply 20, posted (10 months 2 weeks 4 days 17 hours ago) and read 3613 times:

Quoting Starlionblue (Reply 19):
Loved that movie when I was a kid!

I still do ! Mind you, I'm still a kid when it comes to these movies...   



One "oh shit" can erase a thousand "attaboys".
User currently offlinetb727 From United States of America, joined Jun 2005, 1586 posts, RR: 9
Reply 21, posted (10 months 2 weeks 4 days 9 hours ago) and read 3575 times:

So you say something in the cockpit and it gets done? They've had that technology for decades, it is called an FE! Oh wait, they don't make those models anymore.


Too lazy to work, too scared to steal!
User currently offlinesturmovik From India, joined May 2007, 509 posts, RR: 0
Reply 22, posted (10 months 2 weeks 4 days 3 hours ago) and read 3542 times:

When we move to single pilot operations (and its a when, not if) a lot of the functions of the FO will be part of the automation. One of the possible approaches being thought about for flight deck design in this case is a 'virtual pilot' or 'automation assistant' of sorts. IMO, voice recognition will most likely be a part of the flight deck in such a scenario. It certainly is being actively researched.

Quoting Mir (Reply 16):
I've never had a problem manipulating controls in turbulence. I've never tried to use a cockpit touchscreen in turbulence, but I would very much hope that the avionics designers thought about that during the design process.

In my opinion after having tried out some early concepts in lab conditions(not actual flight), the first generation of cockpit touchscreens will most likely be crap, but not for the reasons you might think. Tons of concepts have been developed to address the problems of hand stabilization and input error reduction when using touchscreens in turbulence and high vibration environments, and some of them actually work.

Quoting tb727 (Reply 21):
So you say something in the cockpit and it gets done? They've had that technology for decades, it is called an FE! Oh wait, they don't make those models anymore.

    



'What's it doing now?'
User currently offlineHAWK21M From India, joined Jan 2001, 31679 posts, RR: 56
Reply 23, posted (10 months 2 weeks 2 days ago) and read 3417 times:

Quoting comorin (Reply 11):

Do you have speech on your smartphone, does it work well?

Exactly.....I use an Iphone and a voice activated handsfree...and it seems to respond to an american or east asian accent, not an Indian one  



Think of the brighter side!
User currently offlinesccutler From United States of America, joined Jan 2000, 5505 posts, RR: 28
Reply 24, posted (10 months 2 weeks 1 day 5 hours ago) and read 3368 times:

Handy product, the VoiceFlight," a speech-driven device that loads a flight plan into Garmin 430/530 navigator.

Not much call for it here in Texas, where most clearances are simple, usually a DP, a waypoint or two any a STAR, or even direct; but in the northeast, lots of Victor Airways with lots of twists, turns and waypoints, a voice-activated clearance-entry system can save a lot of knob-twisting and button-pushing.

Www.VoiceFlight.com



...three miles from BRONS, clear for the ILS one five approach...
User currently offlineMax Q From United States of America, joined May 2001, 4441 posts, RR: 19
Reply 25, posted (10 months 1 week 6 days 8 hours ago) and read 3312 times:

Quoting tb727 (Reply 21):
So you say something in the cockpit and it gets done? They've had that technology for decades, it is called an FE! Oh wait, they don't make those models anymore.

Yes, I definitely prefer that TB !


Best wishes



The best contribution to safety is a competent Pilot.
User currently offlineCaptainKramer From United Kingdom, joined Feb 2012, 225 posts, RR: 0
Reply 26, posted (10 months 1 week 4 days 13 hours ago) and read 3051 times:

I remember reading an article a while back, it was in either Aviation Week or Flight International that an Australian Software Firm, Adacel was developing a Cockpit Voice Recognition System that may find it's way onto the JSF F-35.

Adacel is also briefly mentioned in a article on the internet, regarding Australian Firms involved in the JSF. You can Google Wikipedia and there is a brief rundown of the company.


User currently offlineplebbin From Canada, joined Jul 2013, 10 posts, RR: 0
Reply 27, posted (10 months 1 week 4 days 11 hours ago) and read 3065 times:

After skimming through the comments, it seems most of the comments are directed at using voice to control the aircraft. What I haven't heard is how voice recognition could be used in a passive way to monitor the pilots situational awareness.

Lets take the Asiana Flight 214 accident for example. As the airspeed declined below the target airspeed, none of the pilots called out the airspeed until it was too late. The voice recognition could have caught the lack of anyone speaking up and altered the pilots. Otherwise, the system would be silent if a pilot did notice and called out the airspeed.

Of course, the system would need to be improved to eliminate annoying false-positives. Even more importantly, human factors designers would need to ensure that pilots do not rely too much on this system -- in the same way pilots rely on autopilot and autothrottle.

I'm not convinced that voice recognition is for the flight deck but it would be interesting to see what the research finds out.


User currently offlinen6238p From United States of America, joined Jan 2006, 501 posts, RR: 0
Reply 28, posted (10 months 1 week 4 days 10 hours ago) and read 3051 times:

Call up Lockheed Martin flight service and tell me how far we've come with voice recognition software.

BRIEFER

im sorry can you repeat that

BRIEFER

im sorry can you repeat that

BRIEFER

what state will you be departing from

ILLINOIS

i think you said Nevada, is this correct?

NO



To actively root against anybody is just low, and I hope karma comes back at you with a vengeance
User currently offlineStarlionblue From Greenland, joined Feb 2004, 17017 posts, RR: 67
Reply 29, posted (10 months 1 week 4 days 10 hours ago) and read 3046 times:

Quoting plebbin (Reply 27):
The voice recognition could have caught the lack of anyone speaking up and altered the pilots.

I know you meant "alerted", but your phrasing is much funnier. 



"There are no stupid questions, but there are a lot of inquisitive idiots."
User currently offlinejetwet1 From United States of America, joined Jun 2007, 211 posts, RR: 0
Reply 30, posted (10 months 1 week 4 days 6 hours ago) and read 3017 times:

At this point in time my car cannot even call the correct person when I ask it to, there is no way I would step foot on a plane that has voice recognition on the flight deck.

User currently offlineStarlionblue From Greenland, joined Feb 2004, 17017 posts, RR: 67
Reply 31, posted (10 months 1 week 4 days 6 hours ago) and read 3014 times:

Quoting jetwet1 (Reply 30):
At this point in time my car cannot even call the correct person when I ask it to, there is no way I would step foot on a plane that has voice recognition on the flight deck.

I don't think this argument holdswater. You might not step into a car with an autopilot either but you have no problem getting on a plane with one. A normal car costs $20000-30000. A normal airliner costs $30000000-150000000. There is a bit of a difference in the quality and quantity of engineering, certification and workmanship.

Heck, I trust a $400000 Cessna to follow a course using an autopilot that looks like a 70s car radio. It is accurate within a few feet. If the FAA, Boeing, Airbus and the airlines tell me voice recognition is safe in the cockpit, I'll be fine with it.

[Edited 2013-10-15 21:24:50]

[Edited 2013-10-15 21:25:04]

[Edited 2013-10-15 21:29:51]


"There are no stupid questions, but there are a lot of inquisitive idiots."
User currently offlinepilotpip From United States of America, joined Sep 2003, 3150 posts, RR: 11
Reply 32, posted (10 months 1 week 4 days 5 hours ago) and read 2998 times:

Quoting Starlionblue (Reply 6):
Quite. I am reminded of the F/A-18, where if the plane goes out of control arrows on the displays tell the pilot where to position the stick. No need for voice commands.

The ERJ-170/190 already has something like this and I would imagine others do as well. The flight path vector turns off and red arrows appear in the direction to correct when pitch attitudes are exceeded.



DMI
User currently offlineStarlionblue From Greenland, joined Feb 2004, 17017 posts, RR: 67
Reply 33, posted (10 months 1 week 4 days 5 hours ago) and read 2992 times:

Quoting pilotpip (Reply 32):
Quoting Starlionblue (Reply 6):
Quite. I am reminded of the F/A-18, where if the plane goes out of control arrows on the displays tell the pilot where to position the stick. No need for voice commands.

The ERJ-170/190 already has something like this and I would imagine others do as well. The flight path vector turns off and red arrows appear in the direction to correct when pitch attitudes are exceeded.


Fair point. Same on the G1000. However if memory serves the F/A-18 takes it further. Instead of just making red chevrons or arrows on the artificial horizon section of the display, it wipes everything on the screens and replaces with arrows. A bit harder to miss even if your eyeballs are being squished at 7 Gs I suppose.

Can't find a picture unfortunately but here's the G1000.



[Edited 2013-10-15 22:30:50]


"There are no stupid questions, but there are a lot of inquisitive idiots."
User currently offlineHAWK21M From India, joined Jan 2001, 31679 posts, RR: 56
Reply 34, posted (10 months 1 week 3 days 2 hours ago) and read 2855 times:

Quoting Starlionblue (Reply 31):

Lets put it this way....how many pax will be ok travelling on a plane with no pilot monitoring/flying.



Think of the brighter side!
User currently offlineStarlionblue From Greenland, joined Feb 2004, 17017 posts, RR: 67
Reply 35, posted (10 months 1 week 3 days 2 hours ago) and read 2850 times:

Quoting HAWK21M (Reply 34):
Quoting Starlionblue (Reply 31):

Lets put it this way....how many pax will be ok travelling on a plane with no pilot monitoring/flying.

Today, very few. In 100 years, who knows?



"There are no stupid questions, but there are a lot of inquisitive idiots."
User currently offlinesturmovik From India, joined May 2007, 509 posts, RR: 0
Reply 36, posted (10 months 1 week 2 days 2 hours ago) and read 2771 times:

Quoting Starlionblue (Reply 35):
Today, very few. In 100 years, who knows?

True. I may have mentioned this story before, but once when we were trying out some touchscreen prototypes (which do not support multi touch) with some pilots, we had an interesting experience. A coupla old retired airline captains wouldn't even look at it, let alone participate in the study, claiming that they'd never work in a cockpit environment. These guys had flown Avro748s, then 732s, 738s and A320s, and hated the Airbus automation. We took the prototypes to a coupla guys fresh outta flight school, and the first thing they did was to try and pinch-to-zoom the moving map.
Attitudes to technology change with time. That will be the case with voice recognition, single pilot ops and pilotless ops. There will be a lot of kicking and screaming, but it'll happen eventually is my guess.

Quoting n6238p (Reply 28):
Call up Lockheed Martin flight service and tell me how far we've come with voice recognition software.

I've seen prototypes of a company developing voice recognition technology for flight deck use, and it was incredibly good. We had people with thick Indian accents try it out, and it worked with an impressive rate of accuracy. I know that's not a statistically valid case, but our American colleagues could barely understand these same guys on our weekly teleconferences, whereas the software understood what they were saying. I thought that was pretty amazing.



'What's it doing now?'
User currently offlineautothrust From Switzerland, joined Jun 2006, 1595 posts, RR: 9
Reply 37, posted (10 months 1 week 1 day 23 hours ago) and read 2751 times:

The Typhoon has a very advanced speech recognition.

You can give multiple orders like changing frequency of radio/IFF/Routes or do manipulation of the Laser Designator Pod and its picture /creation of a waypoint at a point of interest with only two commands.

Also it features speaker independent voice recognition system based on common databases (e.g. “British English”, “American English”, “Spanish English” etc., no pilot templates)
Large vocabulary

Significant tolerance against variations in the acoustic signal (e.g. ability to cope with pilot breathing)
Voice level variability
The ability to cope with microphones with different frequency responses
The ability to cope with changing environmental conditions (cockpit noise)

Such a system could be implemented in a airliner without having to worry about safety, touogh the costs are massive.



“Faliure is not an option.”
User currently offlineHAWK21M From India, joined Jan 2001, 31679 posts, RR: 56
Reply 38, posted (10 months 1 hour ago) and read 2441 times:

Quoting sturmovik (Reply 36):
I know that's not a statistically valid case, but our American colleagues could barely understand these same guys on our weekly teleconferences, whereas the software understood what they were saying. I thought that was pretty amazing

This is interesting......and amazing too.



Think of the brighter side!
User currently offlinewoodreau From United States of America, joined Sep 2001, 1032 posts, RR: 6
Reply 39, posted (9 months 4 weeks 1 day 6 hours ago) and read 2333 times:

To be honest, I didn't pay very much attention to this thread, until I read the ups 6 crash report

After reading the crash report, and rereading this thread, I'm thinking this is one instance where voice recognition might have helped to reduce pilot workload due to a smoke filled cockpit.

Because the cockpit was filled with smoke, the pilot was unable to retune his radios, and unable to see any of his instruments or look outside the window. having a voice recognition retune radios, nav radios, set heading and or speed bugs or fms entries would have been really helpful in this instance.

Granted there was more going on and even voice recognition would not have saved the plane and crew.

Then reading about google glass in the paper, if there was a google glass/HUD set up inside the smoke goggles/o2 mask displaying HUD symbology on the glass lens inside the goggles paired with voice recognition maybe that would have helped the FO flying single pilot stay oriented as to where he was

Anyways a very expensive solution to very rare instances that don't come up very often.



Bonus animus sit, ab experientia. Quod salvatum fuerit de malis usu venit judicium.
User currently offlineStarlionblue From Greenland, joined Feb 2004, 17017 posts, RR: 67
Reply 40, posted (9 months 4 weeks 1 day 6 hours ago) and read 2322 times:

Quoting woodreau (Reply 39):
Then reading about google glass in the paper, if there was a google glass/HUD set up inside the smoke goggles/o2 mask displaying HUD symbology on the glass lens inside the goggles paired with voice recognition maybe that would have helped the FO flying single pilot stay oriented as to where he was

Anyways a very expensive solution to very rare instances that don't come up very often.

Helmet mounted HUDs are already in use by the military. This enables them to stay in the field of view even if the pilot turns his/her head. It also enables a larger display relative to the field of view. Methinks this technology would be a good thing in a smoke mask. There is the issue with rare instances but the cost of technology tends to plummet with time. Give it a decade or two...



"There are no stupid questions, but there are a lot of inquisitive idiots."
User currently offlinesturmovik From India, joined May 2007, 509 posts, RR: 0
Reply 41, posted (9 months 3 weeks 5 days 1 hour ago) and read 2101 times:

Quoting Starlionblue (Reply 40):
Helmet mounted HUDs are already in use by the military.

There's also a lot of research happening on cheaper near-to-eye displays, even for the civilian space.

Quoting Starlionblue (Reply 40):
Give it a decade or two...

That's the frustrating part, the waiting.  



'What's it doing now?'
User currently offline3rdGen From Bahrain, joined Jul 2011, 235 posts, RR: 0
Reply 42, posted (9 months 3 weeks 4 days ago) and read 2050 times:

For me the odds are severely stacked against voice recognition.

Firstly you have to understand that as great as computers and technology is for the cockpit, all pilots know that you only get good output when you insert good input. Put in the correct waypoints and the FMS and ensure that all the air data is correct and you'll get the plane to fly itself half way across the world. Fantastic.

Input the wrong waypoints or get false air data and what seems so amazing is now a death trap, e.g. AA 767 Cali. AF 447

So yes speech recognition would be fantastic, IF and only IF the system could accurately understand what the pilots were asking it to do. That is where things get complicated. Whether you are German, Indian, Arab, English etc. when you push a button, the computer knows exactly what you want. With the different accents, speech inflections, and background noise, speech recognition will be an enormously challenging task. Added to the challenge is the fact that due to regulation the system would have to be flawless, perhaps a 99.9999% accuracy rate would have to be required before the system could be certified.

The system would also have to have a gigantic database of every possible speech accent and pattern to work. Due to differences in the number and types of letters in various languages many English words are mispronounced by many non native English speakers. E.G. Germans/Indians/Iranians swap the sound of V and W, Arabs don't have the letter P in their language so substitute it with a B sound, etc. there are many examples like this and this adds to the complexity of speech recognition.

Yes the system has been tried successfully on military equipment, but there are two major differences between military and commercial ventures. Firstly, military equipment goes to specific countries and is usually operated by natives of that country, so speech recognition can be tailored to each customer. With commercial aviation there's no telling who is going to be stepping into the cockpit, so the system has to be ready for use with any possible accent or dialect, right off the bat. The second issue is that commercial transportation has much more stringent rules for certification that he military. What might be acceptable for the military would not come close to being acceptable when you have hundreds of paying passengers on board.


User currently offlinerwessel From United States of America, joined Jan 2007, 2342 posts, RR: 2
Reply 43, posted (9 months 3 weeks 3 days 6 hours ago) and read 1988 times:
Support Airliners.net - become a First Class Member!

Quoting 3rdGen (Reply 42):
So yes speech recognition would be fantastic, IF and only IF the system could accurately understand what the pilots were asking it to do. That is where things get complicated. Whether you are German, Indian, Arab, English etc. when you push a button, the computer knows exactly what you want. With the different accents, speech inflections, and background noise, speech recognition will be an enormously challenging task. Added to the challenge is the fact that due to regulation the system would have to be flawless, perhaps a 99.9999% accuracy rate would have to be required before the system could be certified.

Humans can't hit buttons with that level of reliability either.


User currently offlineStarlionblue From Greenland, joined Feb 2004, 17017 posts, RR: 67
Reply 44, posted (9 months 3 weeks 2 days 21 hours ago) and read 1966 times:

Quoting rwessel (Reply 43):
Humans can't hit buttons with that level of reliability either.

Quite. A well trained flight crew member makes one error in around a thousand actions. And yet planes do not routinely fall out of the sky because of it.



"There are no stupid questions, but there are a lot of inquisitive idiots."
User currently offline3rdGen From Bahrain, joined Jul 2011, 235 posts, RR: 0
Reply 45, posted (9 months 3 weeks 1 day 1 hour ago) and read 1863 times:

Quoting rwessel (Reply 43):

Sure but it's the human that makes the mistake not the system. I.e. The human hits the wrong button but the system processes the command correctly, the system doesn't register the input incorrectly. But this is the issue with voice recognition, the human can say the right thing but the system will interpret it incorrectly. It's one thing for the technology to work correctly and for the pilot to make a mistake in using it. It's another thing for the pilot to do the right thing and for the system to make a mistake, the latter is unacceptable for commercial aircraft.


User currently offlineMir From United States of America, joined Jan 2004, 21557 posts, RR: 55
Reply 46, posted (9 months 3 weeks 15 hours ago) and read 1838 times:

Quoting Starlionblue (Reply 44):
A well trained flight crew member makes one error in around a thousand actions. And yet planes do not routinely fall out of the sky because of it.

True, however there is still obviously a need to monitor anything automated, including whatever inputs would be made by voice. That will necessitate the pilot taking his attention away from whatever he's doing to confirm that the software is going to do what he wants it to do for at least a brief period of time. That would seem to negate the chief benefit of voice recognition: not having to look away to push buttons.

-Mir



7 billion, one nation, imagination...it's a beautiful day
User currently offlinerwessel From United States of America, joined Jan 2007, 2342 posts, RR: 2
Reply 47, posted (9 months 3 weeks 6 hours ago) and read 1774 times:
Support Airliners.net - become a First Class Member!

Quoting 3rdGen (Reply 45):
Sure but it's the human that makes the mistake not the system. I.e. The human hits the wrong button but the system processes the command correctly, the system doesn't register the input incorrectly. But this is the issue with voice recognition, the human can say the right thing but the system will interpret it incorrectly. It's one thing for the technology to work correctly and for the pilot to make a mistake in using it. It's another thing for the pilot to do the right thing and for the system to make a mistake, the latter is unacceptable for commercial aircraft.

I'll grant you that there's a different psychology associated with the two types of errors, but there's not a difference if you're interested in the safety of the flight. Errors happen, and they need to be contained and dealt with, whatever the source. At the end of the day the question is does this make the flight safer, more reliable, etc. That question cannot be answered at less than a system level.

Quoting Mir (Reply 46):
True, however there is still obviously a need to monitor anything automated, including whatever inputs would be made by voice. That will necessitate the pilot taking his attention away from whatever he's doing to confirm that the software is going to do what he wants it to do for at least a brief period of time. That would seem to negate the chief benefit of voice recognition: not having to look away to push buttons.

And there are ways to deal with that. A voice confirmation, a message on the HUD. No different than the PF calling for "flaps 40", and hearing a "flaps 40" confirmation from the PFN.


Top Of Page
Forum Index

Reply To This Topic State Of Speech Recognition Tech In The Cockpit?
Username:
No username? Sign up now!
Password: 


Forgot Password? Be reminded.
Remember me on this computer (uses cookies)
  • Tech/Ops related posts only!
  • Not Tech/Ops related? Use the other forums
  • No adverts of any kind. This includes web pages.
  • No hostile language or criticizing of others.
  • Do not post copyright protected material.
  • Use relevant and describing topics.
  • Check if your post already been discussed.
  • Check your spelling!
  • DETAILED RULES
Add Images Add SmiliesPosting Help

Please check your spelling (press "Check Spelling" above)


Similar topics:More similar topics...
Why Are There Still Levers In The Cockpit posted Tue Dec 18 2012 01:56:51 by 456
IPads In The Cockpit? posted Tue Dec 13 2011 20:23:39 by btfarrwm
'Busyness' In The Cockpit posted Thu Jul 7 2011 18:36:45 by bristolflyer
Validity Of Overseas Pilot’s Certificates In The U posted Fri Aug 27 2010 08:19:52 by lenbrazil
What Is This Device In The Cockpit For? posted Sat Jan 16 2010 16:00:46 by JAGflyer
Paintbrush In The Cockpit? posted Tue Nov 10 2009 15:47:18 by Flanders
Pilots & Guns In The Cockpit posted Wed Dec 3 2008 20:42:38 by HighFlyer9790
Do Pilots Use The Handholds In The Cockpit? posted Tue Aug 8 2006 17:08:56 by MechEngineer
Smoke In The Cockpit posted Mon May 22 2006 09:21:09 by AirWillie6475
Problem In The Cockpit - What Would You Do? posted Fri Jan 6 2006 22:02:10 by Julesmusician

Sponsor Message:
Printer friendly format