I just tried P2ATC and I am running into problems with voice recognition. The first couple of phrases work fine and then the number two is recognized repeatedly as four. I tried training with a call sign including a two, but no improvement. And sometimes a completely different sentence is recognized.
My system is not the latest one (Intel I5-6600K) , so it may be a performance issue. The GPU is used 100% by FS2020.
I played with settings in FS2020 using minimum settings, but no improvement.
Any other ideas what to do?

A typical example is shown below with the bold fails at the end:

     Info: Simulator Connected
     Info: Recognition Engine Installed
     Info: Grammars Loaded
     Info: ATIS Freq: 128.65
     Info: Clr Del Freq: 126.75
    Pilot: PAINE Ground;  Cessna Hotel Uniform Hotel;  Request Departure Runway One Six Left;
      ATC: Cessna Hotel Uniform Hotel Departure on runway One Six Left is approved. Contact Ground on One Two One Point Eight for taxi instruction.  enjoy
Suggested: ~Ground on One Two One Point Eight Cessna Hotel Uniform Hotel
    Pilot: Ground on One Two One Point Eight;  Cessna Hotel Uniform Hotel;
    Pilot: PAINE Ground;  Cessna Hotel Uniform Hotel;  Request Taxi Four Runway One Six Left;
      ATC: Cessna Hotel Uniform Hotel Taxi to Runway One Six Left Via taxiways Golf, Golf One,  Hold Short Runway One Six Left
Suggested: ~ Taxi to Runway One Six Left Via taxiways Golf, Golf One,  Hold Short Runway One Six Left Cessna Hotel Uniform Hotel
    Pilot: Taxi to Runway One Six Left via Taxiways Golf Golf One Hold Short Runway One Six Left Cessna Hotel Uniform Hotel;
    Pilot: PAINE Ground;  Cessna Hotel Uniform Hotel;
      ATC: Cessna Hotel Uniform Hotel This is PAINE Ground on One Two One Point Eight Please repeat your request.
Suggested: Check your frequency and your request and try again. Cessna Hotel Uniform Hotel
    Pilot: PAINE Ground;
      ATC: Aircraft calling Ground say again with your call sign.
Suggested: Repeat your request with a proper call sign Cessna Hotel Uniform Hotel
    Pilot: PAINE Ground Cessna Hotel Uniform Hotel;  Say Tower Frequency for Cessna Hotel Uniform Hotel;
    Pilot: PAINE Ground Cessna Hotel Uniform Hotel Say Altimeter;
      ATC: Cessna Hotel Uniform Hotel Altimeter is Three Zero Zero One at SNOHOMISH COUNTY
    Pilot: PAINE Tower Cessna Hotel Uniform Hotel;  Request Departure Runway One Six Left;
      ATC: Cessna Hotel Uniform Hotel Departure on runway One Six Left is approved. Contact Ground on One Two One Point Eight for taxi instruction.  enjoy your morning
Suggested: ~Ground on One Two One Point Eight Cessna Hotel Uniform Hotel
    Pilot: Contact Ground on One Two One Point Eight Cessna Hotel Uniform Hotel;
    Pilot: PAINE Ground Cessna Hotel Uniform Hotel;  Request Taxi Four Runway One Six Left for V F R Departure;
      ATC: Cessna Hotel Uniform Hotel Taxi to Runway One Six Left Via taxiways Golf, Golf One,  Hold Short Runway One Six Left Contact Tower on One Two Zero Point Two
Suggested: ~Tower on One Two Zero Point Two Cessna Hotel Uniform Hotel
    Pilot: Cessna Hotel Uniform Hotel;  Contact Tower on One Four Zero Point Two Cessna Hotel Uniform Hotel;
      ATC: Cessna Hotel Uniform HotelTower Frequency is One Two Zero Point Two
Suggested: ~Frequency is One Two Zero Point Two Cessna Hotel Uniform Hotel
    Pilot: Frequency for Slowing to Zero Point Two Seven knots at Uniform Hotel;
      ATC: Cessna Hotel Uniform Hotel Contact Tower on One Two Zero Point Two
Suggested: ~Contact Tower on One Two Zero Point Two Cessna Hotel Uniform Hotel
    Pilot: Contact Tower on One Four Zero Point Two Cessna Hotel Uniform Hotel;
     Info: Pilot2ATC Closing

The Speech Engine doesn't use the GPU...just CPU.

Hi CPU utilization will definitely affect speech recognition.  If it's over about 75-80%, that could be a problem.

Doing the full Speech Training in Grammar Helper as described in the User's Guide will help...but if your CPU or Memory utilization is extremely high, that might also have issues.

I have problem with "four", p2atc understood for instead of four, but never mind, I used say it to pass by 😄

Dave, gundamf91,

thanks a lot for your fast responses. I will check CPU and memory utilization again and try a full speech training. I have done one Ground only.

I did some more testing and saw that in some cases the CPU was at 80% spikes resulting in complete garbage. But in case of the two versus four error the CPU was at 50% and i could not see any spikes. Is there a way in p2atc to receive the voice input from a text file? In that case I could move the voice recognition to my android device and transfer a file.

No, there is no such capability.

However, if you use Grammar Helper you can listen to exactly what the program is hearing by pressing the Replay button after you say a phrase.  This usually helps you figure out if the volume is too low, speech garbled, etc.

Confusing Four and For is quite common since it is the same spoken word.  However, I find it difficult to see how it can confuse four and two unless there is a problem with the reception.

Dave

After another turn of full speech training including some personal grammar phrases, speech recognition improved a lot and is now working as expected. I probably did something wrong here the first time.
Thanks a lot for your help.

The voice input as a text file is still on my wish list and would enable
- using Voiceattack on a different computer for speech recognition
- using Voiceattack as a grammar translator/adding more phrases to P2ATC and use the grammar of Vatsim or any other European phrases.

You probably could implement a simple filewatcher on a changing datestamp and use the text as voice input. It would be great if that could be considered in one of the next updates.

 

On 7/7/2021 at 3:27 AM, huha001 said:

 

The voice input as a text file is still on my wish list and would enable
- using Voiceattack on a different computer for speech recognition
- using Voiceattack as a grammar translator/adding more phrases to P2ATC and use the grammar of Vatsim or any other European phrases.

 

 

Please explain further. I've used voice attack for Elite Dangerous.

gundamf91, if Dave could implement a voice input text file (or alternatively a text input box in P2ATC where you can enter the voice text or alternatively an UDP listener for voice input at a specific port) i could think of the following used cases:

Case1: Speech recognition with VoiceAttack on a different computer
Create a profile with VoiceAttack on a different computer, which just consists of a dummy "Spoken Command" which is used as unrecognized action in the settings of VoiceAttack. The action is just to dump the recognized voice input into a text file on the P2ATC computer which is then automatically recognized as voice input by P2ATC. The text file must have write attributes for every user in the local network. I would volunteer to create a profile for VoiceAttack if Dave would implement a voice input text file or one of the alternative options in P2ATC. This would even work with the demo version of VoiceAttack.

Case2: Using VoiceAttack as a grammar translator/adding more phrases to P2ATC and use the grammar of Vatsim or any other European phrases
In that case the profile of Case 1 would be extended by "Spoken Commands" in Voice Attack to use your own grammar phrases and the Action in VoiceAttack is to trigger the desired action in P2ATC by outputting the exact P2ATC phrase in the text file. Each spoken command would correspond to a single phrase for which you want to have a specific wording for (special ATC phrases, Vatsim, different language, ...)

Case3: Use Cortana for voice recognition
I believe it would also be possible with some small scripts to start Cortana in dictation mode and let her do the voice recognition task. The recognized text would then be transferred to the P2ATC computer into a text file.

For Case 1,2,3 the PTT sender/receiver app of highflyer525 would be used to bring the PTT joystick command to the client computer, and the command is used to bring VoiceAttack into "listening" mode.

It would also allow to use P2ATC together with the great VoiceAttack profiles of VileIntent604 at https://flightsim.to/file/5247/msfs-plug-in-for-voice-attack and  jamescl604 at https://github.com/jamescl604/MSFSVoiceAttackPlugin.
As I am still having sometimes performance issues on my P2ATC computer this feature would be in my mind a great enhancement for P2ATC.