Improving Voice Recognition Consistency

IMPORTANT.  You will need to be using RightBooth version 7.6.28 (or newer) in order to implement the solution described in this article.

I had an interesting email chat with one of my RightBooth users about the consistency of the Voice Recognition feature.  He was simply wanting his event users to say the word 'cheese' in order to trigger the taking of a photo.

So he created a photo capture event and on his event Start screen he added the word 'cheese' as a label item having the action 'Next screen', and then he turned on the Voice Recognition feature.

But when playing the event he was finding that the spoken word 'cheese' was not always being recognised as such, and this led to our email chat.

On testing this myself, I discovered that when speaking the word 'cheese', the recognition engine was occasionally (and incorrectly) recognising my spoken word as one of the following similar sounding words:

• trees
• he's
• she's
• jeez

Now this recognition inconsistency is actually inherent in the Microsoft Voice Recognition system and it makes RightBooth appear to ignore quite a few spoken words.

So here is my suggestion to improve the reliability:

• On the Event Start screen, add a label item and enter the word 'cheese' AND all the similar sounding words that the engine incorrectly recognises. 
• Note that you must add each word on a separate line in the label item (see image below). 
  
• Set the label item's click action as 'Next screen'.
• Finally, move the label item off the edge of the screen so that it isn't visible to the user when the event plays, like this...

​Job done!

Now when any user says 'cheese' and it is incorrectly recognised as one of the similar sounding words, RightBooth will still accept the word as a voice command and will progress to the next screen as intended, effectively making the voice recognition more reliable and consistent.