I agree.barnabas1969 wrote:Steve,
I think the option to mute would be acceptable. It should be optional however. The user should have the ability to choose mute, dip volume, or neither.
I'll make another reply here with some detailsCan you provide more information about the custom commands?
I can't say for sure as we don't know how much work is involved untill we fasten down all the features that would go into it but there is an alternate product out there called just KinectIf you release a full version of Amulet Voice Kinect, when do you expect it to be ready, and how much do you expect it to cost?
for Mediacenter , it sells for $6.99 so it would need to be in that ball park.
again I agree, I can even include a couple of more postures and have them trigger generic events that are customisable by the user.EDIT: It would be even better if the user had the ability so that when a hand is raised, the software could send a custom command (e.g. to the AVR via RS-232/Ethernet/whatever) to reduce the volume. And then, when the hand is lowered, send another custom command.
However I am limited by the number of postures that I can include as they are not like gestures detected by the Kinect skeletal recogniser, the postures need to be very dependable and not trigger false
positives. This is because the user/s are normally seated in the field of view of the Kinect and the last thing you want to happen is if someone just stretches because maybe their tired or something and
the lights come on or the volume dips, that would get annoying real quick.
The postures at the moment give a really great user experience because they are accurate and dont give positives, that cant be compromised in any way. The same can't be said for the skeletal recogniser.
I experimented with Kinect's built-in skeletal recogniser to recognise gestures early on in the project, but because of the way it works, it will often lose tracking and jitter a lot , when this happens the user
is seen as a twisted mess of bones, causing false positives. I found it impossible to get the level of reliability needed. So I abandoned use of the built-in skeletal recogniser and instead went with detecting
postures using a technique from OpenCV called Haarcascades. I think were the only people using this with the depth feed so far , but I expect more projects to use the technique in the future.
theres more details about it all here
http://www.amuletdevices.com/blog/index ... -down.html
if you read both part one and two of the blog there are videos etc..
Steve.