[Kde-accessibility] Review Request 111267: Automatic detection of SNR in Simon without having to go through the tool for training

[Peter Grasch]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://git.reviewboard.kde.org/r/111267/#review35254
-----------------------------------------------------------



simonlib/simonsound/devicevolumewidget.cpp
<http://git.reviewboard.kde.org/r/111267/#comment25828>

    This means that the monitored "window" of voice activity is 10*<buffer size of level calculation>
    
    The frequency of the calls to deviceReportedLevel() is almost arbitrary and depends on hardware (among other things).
    
    Let's assume, for example, that your hardware has an 200ms buffer (low but possible) - then you'd monitor 2000 ms or 2 seconds of speech data. If the device uses a large, 1 second buffer, you'd be looking at 10 seconds.
    
    I'd say around 10 seconds would be reasonable (I'd try it out and see how that looks) but there's no way to make the window size "10 seconds" by just adjusting the max length of listOfLevels. You'll have to keep track of the stream time (represented by the argument "time").
    
    As the buffer size may change mid-stream, I'd recommend you store the stream time in addition to the level (e.g. QPair) and remove everything older that is now older than 10 seconds (remember, you get the *current* stream time as argument).


- Peter Grasch


On June 28, 2013, 7:45 p.m., Sanjiban Bairagya wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://git.reviewboard.kde.org/r/111267/
> -----------------------------------------------------------
> 
> (Updated June 28, 2013, 7:45 p.m.)
> 
> 
> Review request for KDE Accessibility and Peter Grasch.
> 
> 
> Description
> -------
> 
> This patch attempts to make an individual be able to check if the signal-to-noise-ratio detected by his/her mic is sufficiently high or not just by talking infront of the mic, without having to do any training separately.
> 
> 
> Diffs
> -----
> 
>   simonlib/simonsound/devicevolumewidget.h 06f4e0f 
>   simonlib/simonsound/devicevolumewidget.cpp 6163aa8 
> 
> Diff: http://git.reviewboard.kde.org/r/111267/diff/
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Sanjiban Bairagya
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-accessibility/attachments/20130628/b701f5c9/attachment-0001.html>