Board index FlightGear Support Tools ATC-pie

TTS and speech recognition on Windows

ATC-pie is a radar air traffic control simulation program for the FlightGear multi-player network.

TTS and speech recognition on Windows

Postby OndrejBakan » Thu Jul 19, 2018 9:52 pm

Hello,

I installed ATC-pie yesterday, it's great piece of software! Hail to the creator, really. I think your software can compare to paid one like "global ATC" even if the solo mode is not a main feature. I control on IVAO network, but I don't have enough experience for vectoring on APP position, so I use this for training. I love that I can try CTR position too. That is not an option for me on IVAO, as I don't have a needed exam.

I run ATC-pie in Python 3.7 venv, these packages are installed:

Code: Select all
Package            Version
------------------ -------
inflect            0.3.1
irc                16.3
jaraco.classes     1.5
jaraco.collections 1.5.3
jaraco.functools   1.20
jaraco.itertools   2.4
jaraco.logging     1.5.2
jaraco.stream      1.2
jaraco.text        1.10.1
more-itertools     4.2.0
pip                10.0.1
pocketsphinx       0.1.15
PyAudio            0.2.11
pypiwin32          223
PyQt5              5.11.2
PyQt5-sip          4.19.12
pyttsx3            2.7
pytz               2018.5
pywin32            223
setuptools         39.0.1
six                1.11.0
tempora            1.13


Unfortunately, I can't get a TTS to work. I hear notifications, etc., but the pilots are not speaking. Can you help me with that? What could be the reason? Even though my Windows language is not English, it's Czech, pyttsx3 does work in python console.

Speech recognition works, BUT... as I am not a native speaker, I get mixed results.

For example, when I say:
"4102, PROCEED DIRECT TANGO UNIFORM MIKE KILO ALFA", I get "4102, PROCEED 98..." It looks like the word "DIRECT" is not expected, because when I say only PROCEED, it works fine. Second example is "4102, DESCEND THREE THOUSAND"... Usually I get "4102, climb 30 000" or "4102, climb 80 000". The best in this situation is to say "descend FL 030", but that's not correct, because TA is 5000 ft here. Funny thing is that when I try to mimic the "american" accent, it gets better.

Is there any way to make the speech recognition better? Like fine tuning or something?

Extracting background files from VATSIM or IVAO sectorfiles doesn't work for me. After extracting IVAO files I get a file with SID and STAR in filename, but it contains TSA and TRA, not procedures. (I made background file by hand, as I initially needed only one airport). IvAc2 files seems to be much better in terms of structure, what about extractor for these?

P.S.:
I have to say it again, this game slash simulator is really great. I would love to see it looking more professional in terms of design, to really mimic a radar screen (Euroscope and/or IvAc2, or even IvAc1 are good examples), because under the hood it's great already to become a rival for global ATC and ATC Pro, etc.
OndrejBakan
 
Posts: 1
Joined: Thu Jul 19, 2018 9:33 pm

Re: TTS and speech recognition on Windows

Postby mickybadia » Mon Jul 23, 2018 11:41 am

Hello,

First of all thank you very much for your encouraging words, and welcome to this forum. I am glad it provides excitement to users like yourself. :-)

OndrejBakan wrote in Thu Jul 19, 2018 9:52 pm:Unfortunately, I can't get a TTS to work. I hear notifications, etc., but the pilots are not speaking. Can you help me with that? What could be the reason? Even though my Windows language is not English, it's Czech, pyttsx3 does work in python console.

Honestly no idea with so little insight. Confirm you at least know of System >> Solo set-up (Alt+F11) and of its bottom right-hand option box?

OndrejBakan wrote in Thu Jul 19, 2018 9:52 pm:Speech recognition works, BUT... as I am not a native speaker, I get mixed results. [...]
Funny thing is that when I try to mimic the "american" accent, it gets better. Is there any way to make the speech recognition better? Like fine tuning or something?


This is unfortunately quite normal. Quoting own answers posted in this topic in response to the same concern:
  • "If not native American, you will be departing from what the default acoustic model is expecting. I personally get things to work with a British accent, but sometimes need to force a few "rrr" sounds or distort my input with other American features to help out."
  • "Yes, you can in theory adjust it entirely to your pronunciation, but be prepared for a little work if you try. As I did not like to serve only Americans (it is only I don't like, not Americans lol), I implemented an option to provide a custom acoustic model for recognition, see solo system settings. If you are ridiculously lucky, you find one for "English with your accent" (or even yet: "... for aviation phraseology") on the internet. But since this is unlikely, the solution is to train your own model. At this point I have to let you read and follow the PocketSphinx documentation on training acoustic models, but the bottom line is that you need a relevant (thus significantly long) set of instructions for which you record your voice."

OndrejBakan wrote in Thu Jul 19, 2018 9:52 pm:Extracting background files from VATSIM or IVAO sectorfiles doesn't work for me. After extracting IVAO files I get a file with SID and STAR in filename, but it contains TSA and TRA, not procedures. (I made background file by hand, as I initially needed only one airport). IvAc2 files seems to be much better in terms of structure, what about extractor for these?

All sector files I have seen out there are messy, cluttered with lines commented out and copy-paste redundancies, so I am not surprised. If you know a little what should go in there and where your SID/STAR procedures are contained in the file, you can check against the "output/bg-extract.err" error log file generated by ATC-pie on extract. It is likely that a single empty line or comment mark ";" is messing with the script; it should be reported in the error log. From memory, I would say the v2 format does not make any difference here.

You can try sharing the source file and describing what you were expecting (e.g. a screen shotof your IVAC radar?) so I can diagnose as well.

OndrejBakan wrote in Thu Jul 19, 2018 9:52 pm:I would love to see it looking more professional in terms of design, to really mimic a radar screen (Euroscope and/or IvAc2, or even IvAc1 are good examples), because under the hood it's great already to become a rival for global ATC and ATC Pro, etc.

If you mean something specific, do not hesitate to propose. If you mean fonts, GUI colours and layout, most of it is customisable (see FAQ, the "settings/colours.ini" file and Qt stylesheets). If you come up with a great app stylesheet, we can make it default. CaptB on this forum was making one for a professional look. Not sure if it has reached the shareable state.
mickybadia
 
Posts: 401
Joined: Tue Sep 24, 2013 9:12 am


Return to ATC-pie

Who is online

Users browsing this forum: No registered users and 0 guests