ATCPro Forums Homepage
Forum Home Forum Home > ATC Pro Forums > ATC Pro Official Support
  New Posts New Posts RSS Feed - Speech Recognition
  FAQ FAQ  Forum Search   Events   Register Register  Login Login

Speech Recognition

 Post Reply Post Reply
Author
Message
Mick View Drop Down
Member
Member


Joined: 21 May 2018
Location: Coburg
Status: Offline
Points: 6
Post Options Post Options   Thanks (0) Thanks(0)   Quote Mick Quote  Post ReplyReply Direct Link To This Post Topic: Speech Recognition
    Posted: 28 May 2018 at 3:46am

I need support with the speech recognition.


Speech recognition works partially and sometimes, after adjusting the microphone position or sensitivity, even poorly.

1). Youtube-Video: 16.05.18 (Assessment speech recognition 75-80% => sensitivity: 81) - Error on these words:
runway (recog: "number"), ILS (recog: "a d f") and x miles from supot


https://www.youtube.com/watch?v=pDkdVjnxsUk

 

2). Youtube-Video: 21.05.2018 (changing the microphone sensitivity from level 80 to 83)
After changing the sensitivity, it takes almost an eternity for speech recognition to improve again.


https://www.youtube.com/watch?v=JE1HCSN6Vx8&t=1s

 

What can I improve?

 

System: Windows 7 Ultimate
ServicePack 1
Processor: Intel(r) Core(TM) i5-2500 CPU@3.30GHz
Installed memory (RAM): 12,0 GB
System type: 64-bit Operating System

 

Headphone: saitek pro flight series headset
https://www.amazon.de/Saitek-PH09-Pro-Flight-Headset/dp/B001EYU1WI

 

Microphone settings:
https://www.youtube.com/edit?o=U&video_id=IVSuS1gFk-I

 

Note: I find that ATC-Pro is the best ATC simulation available.


greetings


Mick

Coburg

Bavaria, Germany

 

 



Edited by Mick - 31 May 2018 at 12:46pm
Back to Top
Tom@FlagMountain View Drop Down
Admin Group
Admin Group


Joined: 22 Oct 2014
Location: Albuquerque, NM
Status: Offline
Points: 509
Post Options Post Options   Thanks (0) Thanks(0)   Quote Tom@FlagMountain Quote  Post ReplyReply Direct Link To This Post Posted: 28 May 2018 at 5:33pm
Mick:
 
The link to the video in your email doesn't work.  Please post to dropbox, or some other online storage location.
 
When you are in an ATCpro session, and have difficulity with SR, please send me the session log file.  File path is C:\programdata\Flag Mountain\ATCpro\system\atcprolog.txt.
 
Would also recommend that you make some audio clips for me to analyze.  Use Audacity, which is a free utility.  Separate each command into its own WAV file.
 
Regards
Tom Murdock
Flag Mountain Software
ATCpro Project Team
Back to Top
Mick View Drop Down
Member
Member


Joined: 21 May 2018
Location: Coburg
Status: Offline
Points: 6
Post Options Post Options   Thanks (0) Thanks(0)   Quote Mick Quote  Post ReplyReply Direct Link To This Post Posted: 29 May 2018 at 1:49am
I corrected the links. A log file is already attached. I still create the audio files.


Back to Top
Mick View Drop Down
Member
Member


Joined: 21 May 2018
Location: Coburg
Status: Offline
Points: 6
Post Options Post Options   Thanks (0) Thanks(0)   Quote Mick Quote  Post ReplyReply Direct Link To This Post Posted: 31 May 2018 at 12:45pm
Audiofiles sent to username Tom @ FlagMountain.

greetings

mick
Coburg
Bavaria, Germany
Back to Top
Mick View Drop Down
Member
Member


Joined: 21 May 2018
Location: Coburg
Status: Offline
Points: 6
Post Options Post Options   Thanks (0) Thanks(0)   Quote Mick Quote  Post ReplyReply Direct Link To This Post Posted: 23 Jan 2019 at 12:49pm
Subject - RE: audio file analysis
Sent - 07 Jun 2018 at 9:25am
Sent by: Mick

Hello Tom,

Thank you for the quick processing of my request.

It is a pity you can not help me.

It can not be that I'm the only user with an accent. Certainly there are also users of the ATCpro application in

the United States or Canada who speak with an accent.

Also, I can record the Google Voice output, which is absolutely free of accents
adjust speed and check the result.

As a further option, I can also bite a linguist friend to record the six wavefiles again.
It would help a lot, if the audio file analysis is done again.

This answers still does not answer the question why speech recognition has been deteriorated,
when the microphone sensitivity is changed. [Change of the level from 81% to 83%]
If it is only the accent, it has a time, in front of the microphone position or
sensitivity was changed, worked.

Receipt:
Video: https://www.youtube.com/watch?v=pDkdVjnxsUk
For the following aircraft identifications:
- SWA338 / Timeline: 6:24 min.
- AMF1351 / Timeline: 9:21 min.
- N449KB / Timeline: 12:52 min.
- AMF1351 / Timeline: 15:55 min.


Same sentence / result of speech recognition:
From 12:24 min. until 12:51 at this video, I repeat the same sentence:
[N449KB cleared ils runway 8 approach].
The pronunciation remains almost identical and my existing accent is also preserved.
In the third attempt, speech recognition recognizes the sentence, despite the accent.

The voice recognition gives the following results:
1. Announcement:
"N449KB cleared ils runway 8 approach"
Result: N449KB hold tcs and then enter final emuxe approved
Comment: Let's assume that the words "Cleared" and "ILS" were not understood.
Why does the SR enter the words "hold tcs and then"? The articulation is not even related ???

2nd Announcement:
"N449KB cleared ils runway 8 approach"
Result: N449KB cleared a d f to runway eight
Comment: Here "runway" understood correctly, only "ils" is wrong.
The command "a d f" does not exist in the voice command list at all.
[Neither in ATCpro Voice Command List NEW!! nor in ATCpro Verbal Command QR Card NEW!!]

3rd announcement:
"N449KB cleared ils runway 8 approach"
Result: N449KB via tafoy enter final emuxe approved
Comment: Here is not a clear recognition - "Let me guess - method ???"

4. Announcement:
"N449KB cleared ils runway 8 approach"
Result: N449KB cleared i l s runway eight approach

Other procedures:
I will record this sentence "N449KB cleared ils runway 8 approach" and recording
play repeatedly at the right time. The realization may be different anyway
results.

Word "runway" to recognize:
Of course, the question arises for me how to pronounce the word "runway" like this, that
it will be recognized correctly. I already have my pronunciation with the Google speech output
and unrecognizable errors detected. It is a simple word and basically not be misunderstood.

The last option, I will test a self-programmed speech recognition in C#
[using System.Speech.Recognition;], which was Windows Speech recognition based,
the text commands from ATC-Pro deposit and check. It will turn out
perhaps, that the commands, despite accent are known.

Here is an excerpt in a result log for the word "runway" from the Windows-Recognition:
(The word is clearly recognized. The articulations are identified behind the word and the time
to be regcognized.)

===================================================================

Data Time:06.03.2018 10:30:49

Grammar(), 00:00:03.4000000: "ground, November 8 5 2 Charlie Mike request to enter runway, 16"
ground,    ground,    gɻa͡ʊnd    00:00:00.3100000 (OneTrailingSpace)
 November   November   novɛmbə    00:00:00.2900000 (OneTrailingSpace)
 8          8          e͡it       00:00:00.0800000 (OneTrailingSpace)
 5          5          fa͡iv      00:00:00.1700000 (OneTrailingSpace)
 2          2          tu         00:00:00.1300000 (OneTrailingSpace)
 Charlie    Charlie    t͡ʃɑɻli    00:00:00.2100000 (OneTrailingSpace)
 Mike       Mike       ma͡ik      00:00:00.2300000 (OneTrailingSpace)
 request    request    ɻikwɛst    00:00:00.4300000 (OneTrailingSpace)
 to         to         tu         00:00:00.1100000 (OneTrailingSpace)
 enter      enter      ɛntə       00:00:00.2800000 (OneTrailingSpace)
 runway,    runway,    ɻʌnwe͡i    00:00:00.3300000 (OneTrailingSpace)
 16         16         wɑnsɪks    00:00:00.4500000 (OneTrailingSpace)
 alt(0.9435087) ground, November 8 5 2 Charlie Mike request to enter runway, 16
===================================================================

Data Time:06.03.2018 10:31:06

Grammar(), 00:00:02.5200000: "enter runway, 16 November 8 5 2 Charlie Mike"
enter      enter      ɛntə    00:00:00.3300000 (OneTrailingSpace)
 runway,    runway,    ɻʌnwe͡i    00:00:00.3100000 (OneTrailingSpace)
 16         16         wʌnsɪks    00:00:00.2200000 (OneTrailingSpace)
 November   November   novɛmbə    00:00:00.3300000 (OneTrailingSpace)
 8          8          e͡it       00:00:00.1300000 (OneTrailingSpace)
 5          5          fa͡iv      00:00:00.2100000 (OneTrailingSpace)
 2          2          tu         00:00:00.1300000 (OneTrailingSpace)
 Charlie    Charlie    t͡ʃɑɻli    00:00:00.2400000 (OneTrailingSpace)
 Mike       Mike       ma͡ik      00:00:00.3200000 (OneTrailingSpace)
 alt(0.9571536) enter runway, 16 November 8 5 2 Charlie Mike
===================================================================



greetings

Mick

(speech recognition is more of a passion than a simple function.)
Back to Top
 Post Reply Post Reply
  Share Topic   

Forum Jump Forum Permissions View Drop Down

Forum Software by Web Wiz Forums® version 11.10
Copyright ©2001-2017 Web Wiz Ltd.

This page was generated in 0.109 seconds.