Anyone using voice recognition for accuracy feedback?

[teach...@gmail.com]

We're in the final stages of perfecting our voice recognition routine that detects "yes" and "no" responses. Our last challenge is the reliability of the recognition routine. We're using Microsoft's SDK 5.1. We'd be interested in hearing about anyone's experience with this type of function. 

Best,

Cathy Smith 

Kent State University

[Michiel Sovijarvi-Spape]

Hi,

This would be incredible... but is it implemented in E-Prime? I think, there are two "systems" commonly used by the rest of us less tech-savvy psychologists.

1. Online: using a keyboard, the experiment supervisor sits in the lab and presses a "y" button for yes and "no" for no. In-cre-di-bly boring, I can tell you. The other system is where the supervisor codes directly whether an answer is accurate or not, but this introduces more errors as people tend to imitate voices (c.f. Spapé & Hommel, Psych Bulletin & Review, 2008!).

2. Offline: using newer E-Prime, you record whatever people say and only later on, go through hours and hours of audio "by hand" (or ear), and score responses. Also boring, but at least you can drink coffee and listen to music while doing so! It does take a bit more time, I find.

 

Is your routine an offline or online automatic system?

Sounds like a dream to many, anyway.

Best,

Michiel

--
You received this message because you are subscribed to the Google Groups "E-Prime" group.
To view this discussion on the web visit https://groups.google.com/d/msg/e-prime/-/-z3dwo2r0ewJ.
To post to this group, send email to e-p...@googlegroups.com.
To unsubscribe from this group, send email to e-prime+u...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/e-prime?hl=en.

[Catherine Smith]

Hi Michiel,

We're doing this online, as we need vocal response feedback that is
comparable to motor response feedback. Our last hurdle is connecting
individual "dictation profiles" for each subject to the recognition
routine.

Subjects will train the system on their unique pronunciation of "yes"
and "no", which is easy to do. This creates a "voice profile" for the
subject. Hopefully with just a little code tweak we can use the unique
profile in the experiment.This will make the response detection much
more reliable. We're also still trying to perfect timing issues, but
we're optimistic that it will work well.

We're happy to hear that this will be useful for others. We'll share
it when we get it running!

Best,
Cathy

--
-----------------------------------------
Dr. Catherine L. Smith
Assistant Professor
School of Library and Information Science
Kent State University
PO Box 5190 – 314V Library
Kent, OH 44242
330-672-2116 (office)