Millisecond Forums

Trouble with Speech recogntion

https://forums.millisecond.com/Topic23985.aspx

By Bruno - 3/8/2018

Hi,
We have some trouble with the speech recognition function in Inquisit5. We use a script of another lab that used it succesfully with Inquisit4. The script and task runs fine. The problem is that the designated responses are not always recognized and the program doesnt trigger (seems to be roughly a third of the responses) .
We use

/ inputdevice = speech
/ validresponse = ("Ja", "Nee", "Yes", "No") 

Any help is appreciated. For instance, we used the Speech Training Wizzard. Could you explain how this works and how best to use it, e.g. is is best to do this for each subject or rather to have multiple subjects do it?
By Dave - 3/9/2018

Bruno - Friday, March 9, 2018
Hi,
We have some trouble with the speech recognition function in Inquisit5. We use a script of another lab that used it succesfully with Inquisit4. The script and task runs fine. The problem is that the designated responses are not always recognized and the program doesnt trigger (seems to be roughly a third of the responses) .
We use

/ inputdevice = speech
/ validresponse = ("Ja", "Nee", "Yes", "No") 

Any help is appreciated. For instance, we used the Speech Training Wizzard. Could you explain how this works and how best to use it, e.g. is is best to do this for each subject or rather to have multiple subjects do it?

Inquisit uses the speech recognition engine that comes with the operating system, it does not have its own. Recognition accuracy generally improves the more and better that engine has been "trained" to a specific speaker, i.e. ideally you would go through some training phase with each individual participant right before the experiment. You can create different recognition profiles for different speakers, under Windows, open Control Panel -> Ease of Access -> Speech Recognition -> Advanced Speech Options.

The training wizard simply calls the operating system's built-in training feature for the recognition engine -- it's the same thing you would via the system's "Speech Recognition" control panel applet. One problem you may run into with your specific code above is that you seem to be mixing languages (Dutch / English). The system's speech engine has some expectations re. its input language, which depends on the system's language settings. You can switch the language via the same Control Panel options described above, but you may have to install additional input languages first (these are available via Windows Update for non-Home editions).