A bit problem with VA

[gundamf91]

Hi everyone,

I've purchased VA a few weeks ago. Now I'm using VA for Elite Dangerous integrated with Astra voice pack from HCS. Sometimes, VA did not recognize command properly, for example when I said 'Hello', VA recognized 'help' or 'enable', VA recognized 'an Anbel'. I'm using headset wireless of Sony, connect to PC via wireless, mic is inside headset not a stick. I'm from Vietnam, maybe my pronunciation is not good, but VA could recognize couple of my commands. So is there any solution for this issue or I have to improve my pronunciation?

And one more thing, sound of VA is loudness than default voice of Elite Dangerous. I've config audio tab (included integrated mode) of VA and improved but not optional. Is there any solution?

p/s: I've searched my issue by google also in this forum but unsatisfactorily, I've been looking for general topic for QA so that I create this topic. If there is any inconvenience, I really apologise.

Thank you very much.

[Pfeil]

Sometimes, VA did not recognize command properly, for example when I said 'Hello', VA recognized 'help' or 'enable', VA recognized 'an Anbel'. I'm using headset wireless of Sony, connect to PC via wireless, mic is inside headset not a stick. I'm from Vietnam, maybe my pronunciation is not good, but VA could recognize couple of my commands. So is there any solution for this issue or I have to improve my pronunciation?

You can try adjusting the "Command Weight" and "Minimum Confidence Level" values on the "Recognition" tab of the VoiceAttack option window.

Also make sure you've trained your speech recognition profile a number of times (three sessions, at least).

Obviously, having a good quality microphone and a quiet environment will allow for more accurate speech recognition, but pronunciation remains essential.

And one more thing, sound of VA is loudness than default voice of Elite Dangerous. I've config audio tab (included integrated mode) of VA and improved but not optional. Is there any solution?

For sounds that are played using VoiceAttack, you should be able to set that using the "Sound File Volume Offset" option on the "Audio" tab, likewise you can set the text-to-speech voice volume using the "TTS Volume Offset" option.

You should also be able to adjust the global volume for all VoiceAttack sounds using the Windows volume mixer, as you would for any other application.

[gundamf91]

Sometimes, VA did not recognize command properly, for example when I said 'Hello', VA recognized 'help' or 'enable', VA recognized 'an Anbel'. I'm using headset wireless of Sony, connect to PC via wireless, mic is inside headset not a stick. I'm from Vietnam, maybe my pronunciation is not good, but VA could recognize couple of my commands. So is there any solution for this issue or I have to improve my pronunciation?

You can try adjusting the "Command Weight" and "Minimum Confidence Level" values on the "Recognition" tab of the VoiceAttack option window.

Also make sure you're trained your speech recognition profile a number of times (three sessions, at least).


Obviously, having a good quality microphone and a quiet environment will allow for more accurate speech recognition, but pronunciation remains essential.

And one more thing, sound of VA is loudness than default voice of Elite Dangerous. I've config audio tab (included integrated mode) of VA and improved but not optional. Is there any solution?

For sounds that are played using VoiceAttack, you should be able to set that using the "Sound File Volume Offset" option on the "Audio" tab, likewise you can set the text-to-speech voice volume using the "TTS Volume Offset" option.

You should also be able to adjust the global volume for all VoiceAttack sounds using the Windows volume mixer, as you would for any other application.

I will try your suggestions. Thank you very much.

[gundamf91]

You can try adjusting the "Command Weight" and "Minimum Confidence Level" values on the "Recognition" tab of the VoiceAttack option window.

Could you please suggest appropriate values for those options? Thank you.

[Pfeil]

Those values would be specific to your setup:

Command Weight
This value is the relative weight of the commands in your profile versus everything else you say. The higher the number, the more likely VoiceAttack will make a, 'best guess' at what you say.
For example, when this value is at maximum (100), your commands have full weight. That means that basically everything you say will be interpreted into a command. If you say, 'flag', and, you have no command called, 'flag', but, the closest match is, 'bag', VoiceAttack will execute the command named, 'bag'.
Note that a high value in this option will probably not be desirable when VoiceAttack's listening is on all of the time (especially if you carry on conversations). However, a high value may be beneficial when using the push-to-talk feature.
Play around with this number to get the right balance for your speaking style.

Minimum Confidence Level
When the Windows speech engine recognizes a phrase, it provides a confidence rating of just how accurate it thinks it was at doing its job.
VoiceAttack will allow you to filter out anything that the speech engine recognizes but does not meet a minimum rating. You can set this value here (from 0 to 100). The higher the number, the more selective VoiceAttack will be about executing commands.
Note that this value can be overridden at both the profile level (on the Profile Options screen) as well as for each individual command (on the Command Add/Edit screen). Any phrase that is recognized but rejected because it falls below the minimum value will show up in the log.
Select the, 'Show confidence level' option to show the speech engine's confidence level in the log for each recognized phrase.

Neither option will "fix" poor recognition by itself, that can only be done by training the speech recognition profile, and providing clear input.

[gundamf91]

Those values would be specific to your setup:

Command Weight
This value is the relative weight of the commands in your profile versus everything else you say. The higher the number, the more likely VoiceAttack will make a, 'best guess' at what you say.
For example, when this value is at maximum (100), your commands have full weight. That means that basically everything you say will be interpreted into a command. If you say, 'flag', and, you have no command called, 'flag', but, the closest match is, 'bag', VoiceAttack will execute the command named, 'bag'.
Note that a high value in this option will probably not be desirable when VoiceAttack's listening is on all of the time (especially if you carry on conversations). However, a high value may be beneficial when using the push-to-talk feature.
Play around with this number to get the right balance for your speaking style.

Minimum Confidence Level
When the Windows speech engine recognizes a phrase, it provides a confidence rating of just how accurate it thinks it was at doing its job.
VoiceAttack will allow you to filter out anything that the speech engine recognizes but does not meet a minimum rating. You can set this value here (from 0 to 100). The higher the number, the more selective VoiceAttack will be about executing commands.
Note that this value can be overridden at both the profile level (on the Profile Options screen) as well as for each individual command (on the Command Add/Edit screen). Any phrase that is recognized but rejected because it falls below the minimum value will show up in the log.
Select the, 'Show confidence level' option to show the speech engine's confidence level in the log for each recognized phrase.

Neither option will "fix" poor recognition by itself, that can only be done by training the speech recognition profile, and providing clear input.

Yesterday, I've tried your first suggestions and it seem to be improved. About training the speed recognition, that is Speech Engine Training utilities in Recognition tab of Voice Attack, is that right?

[Pfeil]

About training the speed recognition, that is Speech Engine Training utilities in Recognition tab of Voice Attack, is that right?

Correct.

[gundamf91]

Correct.

After followed your advises, VA now works almost perfectly. Thank you very much.