[forgot the list carbon copy]

Am Montag, den 26.11.2007, 17:36 +0000 schrieb Daniel Drake:
> Tobias Wolf wrote:
> > I did something real crass (or stupid, not sure) and changed the USB id
> > in aes4000.c to my unsupported model's 08ff:5711 (an AES 3500 as far as
> > I know, Samsung X30 integrated).
> 
> Nice!
> 
> How do you know it's an AES3500? I don't know better, just curious how 
> you reached this conclusion :)

>From Google searches actually. One site demonstrated the Windows .inf
file stating the model.

> I tried to confirm this by looking for the windows driver for this 
> laptop on samsung's site, but I had no luck. I'm assuming the NX30 
> laptop is the one in question (no X30 in the list...)

http://www.samsungpc.com/products/x30/x30_supporteddrivers.htm

The driver is downloadable there.


> > The result is attached. It scanned alright, but the result is striped
> > and blocky. I found that the image format is 128x128. Is there a way to
> > help adding support without knowing how to read raw USB dumps?[1]
> 
> You could submit dumps for others to read :)
> 
> Assuming you can use this sensor under windows, please use this sniffer:
> http://www.pcausa.com/Utilities/UsbSnoop/default.htm

Yeah. Unfortunately I don't have Windows installed. I used to have a
VMWare XP install. That might work too? I would need some procedures for
this. I know there's a diagnostic tool for windows. Would starting the
logger, then starting the diag-tool and doing one scan-pass be enough?

> The image looks a bit blocky because the AES4000 driver includes a 
> magnification of the image (improves image processing performance, 
> temporary workaround for a NBIS bug).
> 
> But yes, even without the magnification it wouldn't look great, and 
> those stripes aren't that nice either. The finger ridges don't seem to 
> line up very well over different stripes.

Might be because AES4000 has 96x96 and not 128x128?

> Also the image is only of a very small area of the finger. Our current 
> image processing code doesn't work that well with such small images 
> (whereas it works brilliantly for images that cover more of the finger). 
> Have you had any success with the enroll and verify example programs (or 
> using the equivalent functionality under fprint_demo)? I'd be interested 
> to see binarized versions and binarized images where the minutiae are 
> plotted (fprint_demo verify interface gives you these controls)

Yes, even with the garbled samples it found some minutiae (24?) and it
registered the finger. The sensor is a small golden square (7x7mm)
between my touchpad buttons. I'll attach a sample print I kept from
Windows.

> I'm pleased that you've got this far though. Even if the image 
> processing performance is not great I would be interested in including 
> this driver to allow others an easy path to work on fixing the other 
> problems. I'd suggest copying aes4000 to aes3500, making your 
> modifications, then submitting a patch. There will be a fair amount of 
> duplicated code but I will reduce this quite soon as there is a good 
> amount of code that all authentec devices can share.

Yeah, only I'm no good at this. The farest I went was changing all
occurences of 96 to 128. That was coarse and didn't work.

I'll check out how USB-sniffing works with VMware ...

--Tobias

<<attachment: right-index.png>>

_______________________________________________
fprint mailing list
[email protected]
http://lists.reactivated.net/mailman/listinfo/fprint

Reply via email to