How can I improve speech recognition using Oculus Rift?
Using Oculus Rift, you can turn off Automatic Gain Control (AGC) as an attempt to better give commands. The microphone in the Oculus Rift CV1 is quite sensitive and letting software constantly change gain on the input is not always ideal. To turn off AGC, search for “Sound” in Windows Start -> Run. Open the Sound settings and switch to the “Recording” tab. Open the “Microphone Rift Audio” settings, here in the “Custom” tab you can turn off AGC. Even though you can’t press Apply, pressing OK stores this setting.
Voice Attacks detects a lot of words that I don’t say and without me saying anything. How can this be improved?
Try the following 5 steps to minimize “false positives” where the speech recognition picks up noise in your environment
- Turn down the volume / gain of your microphone in the Windows Sound settings
- Try enabling “Mute listening when busy” toggle in DRE Launcher -> General in order to mute Voice Attack whenever you are braking in iRacing. This can easily remove some false positives
- You can train your computer to better understand you. Go to Control Panel -> Ease of Access -> Speech Recognition. Click on Train your computer to better understand you. This will make it know how you sound and will be able to better distinguish your voice from noise from eg. steering wheel or other people nearby. Do this training at least twice! Also over time, the recognition gets better as it learns as you talk to it.
- Whenever Voice Attack mistake noise or something you said which wasn’t the correct command you wanted, reply back to DRE what it wrote in the Voice Attack Log. Eg. if recognition picks up “iRating” command from the noise in your room, say “iRating” back to DRE a couple of times. The reason is, that the Windows Speech Recognition supposibly can learn as you go. By learning how it is actually pronounced by you it can better differentiate in the future
- Try playing around with the Voice Attack settings for Recognition. Try the values below (mainly Recognized & Unrecognized Speech Delay, Minimum & Minimum Unrecognized Confidence Level)
Voice attack sometimes doesn’t understand me fully. How can this be improved?
If the speech recognition detects faulty commands, you can optimize it for these exact words or sentences. Go to Start and search for Start Speech Recognition. Right click on the Speech bar on top of your window and click on Open Speech Dictionary. Here you can prevent a word from being detected or add a new word to optimize understanding. Eg. if it has a hard time understanding the word ‘garage’, you can add this as a new word by checking on Record a pronunciation upon Finish. Be sure to turn on listening in the Speech bar before pressing Record.
When I ask for information, all it gives me is ‘None’ or ‘Not set’. Or, I cannot see DRE Licensed despite having Voice Attack full?
Make sure you have initialized The DRE. This is the start up command needed to connect to iRacing first, and it’s called “Initialize The DRE”. To make it easier for you, In Voice Attack open the Edit Profile, and then click on the “Options and overrides for this profile”. In “Execute a command each time this profile is loaded” you can set it to “Initialize The DRE”. This way it will start searching for iRacing every time you open this Profile or start Voice Attack.
Commands are unrecognized when I want to write sentences in the chat, or remind myself about something.
You need to say your sentence immediately after the command. So if you want to write “sorry” in the chat, you need to say “Start writing sorry”, without any pauses. This goes for all your unrecognized commands. Since VA cannot compare your sentence with any command it already knows, it relies on Microsoft Speech Recognition services which is less than optimal. This may be why it often misinterprets your input for these types of commands. It’s recommended to add your own commands in case you often write certain sentences.