Beta using voice recognition thread?

[James_e_Vining]

When you get the Beta ready to test ( I am not rushing you just curious) Will you be posting the info in the Beta thread at the bottom.
I am just wanting to know so I can keep an eye out for it so I can start testing it as soon as you are ready.
Again not rushing you.

[mr_belowski]

Soon - if I get testing time tonight I'll put one up later

[mr_belowski]

i'll put this on the Beta thread soon - here's the latest build with the speech recognition changes.

http://crewchief.isnais.de/CrewChiefV4-4.9.9.0_beta.msi

Note that the default behaviour is to use the existing SRE config (the "Microsoft" SRE). You'll need to go to the properties screen and change 'Prefer Windows speech recogniser' to true (checked), then restart the app.

I'd like to hear if it improves SRE accuracy, particularly after going through the Windows SRE training process (which only takes a few minutes)

[James_e_Vining]

I will get on this as soon as my League race is over Approx 9:30 PDT, Wow I did not expect you to be so fast,
I was not trying to rush you I promise I just wanted to know how to find out when you had something ready.
we are running Darlington and I am having issues 3-4 or I would start working on it right now.

[James_e_Vining]

Ok so I changed my mind I just set up and did the speech recognition training in windows now I am installing CC.
So I will be testing it live basically, first for a bit pre race then during race ( I hope all goes well) worst case is I have to manually select things Like I did pre CC

[James_e_Vining]

So this is what happened, I installed The Beta after doing the voice training and was very happy as it seemed to work perfectly, Black Box (BB) standings, Tires, relative, fuel all worked, What time is it, What is the Strength of Field, How is my fuel, Pitstop Change Right Side Tires, Pitstop add 10 Gallons of fuel.

Then for some reason when race was about to start it want spazz. I would as how is my fuel ( it would recognize) How is my, or black box relative only Black box was there.
I am going to go through a additional 1-2 training sessions and try again later this evening.
I do not understand how it was working perfect then all of a sudden it was missing some words or showing words not even close
im gonna figure for now it is on my end and I need to train it more.
I will update you on progress as soon as I can do more training and run some tests.

[James_e_Vining]

Ok I have great news, After running an additional Voice training session (really not that hard or long) It seems Crew Chief now understands me much better, The only Times it did not react as I expected was when I gave a incorrect command or asked for something I do not believe is mapped (F keys)
I copied and saved the console output but since I was only in session alone and was not running at speed with noise etc I thought I would do another test under race conditions with other drives so I can ask questions about drivers and lap times and so on.
IMHO I really believe the Improvement in the SRE accuracy is incredible.
I will be copying the console data window each time So you can see the progress made (so far I have only done 2 training processes, the original one and then one additional process)

[mr_belowski]

Thanks for the update fella, I've got a bit more of my own testing to do but it won't take long.

You might get a different result in a noisy environment - definitely worth checking

I'll probably push this to the live version soon

[James_e_Vining]

Just ran a practice and a race using it and must say I am very pleased with performance, The Chat option you added Message <>>> etc works also.
I will add the (sorry the name escapes me at moment) data in console window. I have been playing with this for some time and am quite exhausted to be honest .
Only issues I found seem to be user error either i'm not holding button long enough and I can tell not speaking clearly enough but I can work on that.
I think this is a huge addition and thank you for letting me try it out .
There is a whole bunch of commands I have not tried ( mostly as I cannot remember the wording exactly) or had to concentrate on track, Darlington may have been a bad track to test this anyway, talladega may have been better lol
( I had to cut down original size at it was to large the 2 I attached should give you info on it functioning correctly.)

[mr_belowski]

thanks for testing this

I think the free-dictation chat option will have to be removed - the grammar doesn't behave properly and the recogniser ends up using the free dictation grammar instead of the 'proper' grammar even when it shouldn't. I think this is a bug in the library so that feature (which would have been very nice) will have to be disabled

this log snippet shows it:

12:50:24.464 : Recognised : black box in car Confidence = 0.825 Time Elapsed (ms) = 105.9385
12:50:24.464 : chat recognised: black box in car
12:50:24.707 : eric ramirez has entered the pitlane with sane flags
12:50:26.547 : Listening for voice command...
12:50:26.547 : Opened channel - waiting for speech
12:50:26.547 : Getting audio from default device
12:50:28.296 : Purging regular queue
12:50:28.296 : Purging immediate queue
12:50:28.300 : Invoking speech recognition...
12:50:28.300 : Cancelling wait for speech
12:50:28.308 : Recognised : black box car Confidence = 0.955 Time Elapsed (ms) = 7.9979
12:50:28.311 : Sending actions F8
12:50:28.311 : Pressing keys F8

the 'chat recognised' shouldn't be showing up here - this means that the SRE thinks it's heard the magic "Message..." key word so it sends the rest of the command to the dictation recogniser, which is why it doesn't work. You second attempt a couple of seconds later gives the correct result