Hi : Mr Justin Haygood had give us two patches. I try it ,the score had improve .So why didn't you to add them to your codes? May be it will have some harm I don't know? Thank you .
Minimum sample size of PCM stream
Justin Haygood
Apr 6
Hello Mario!
3-5 seconds might be to short, but there's a few changes we did to make it better for short (15-30) content with OTA recognition. I've attached 2 patches that we used that made it work a lot better, however, fingerprints generated aren't compatible with stock libcodegen (however, reference fingerprints created using the patched libcodegen matches sample fingerprints created using the patched libcodegen).
The changes basically involve:
- Increasing density of fingerprints (more data in less time) to increase match rate for short samples
- Reducing false positive rate by skipping over "empty" hashes.
However, we've seen much better match rates (though 3-5 seconds is probably on the short side even then) with some of the proprietary ACR technologies. I highly recommend you take a look at them.
- show quoted text -
- show quoted text - -- You received this message because you are subscribed to a topic in the Google Groups "echoprint" group. To unsubscribe from this topic, visit https://groups.google.com/d/topic/echoprint/J7MQftCfpCM/unsubscribe?hl=en. To unsubscribe from this group and all its topics, send an email to echoprint+...@googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out.
Attachments (2)
changeset_SecondScreen_trunk_Libraries_libcodegen_r51612.diff 774 B View Download
changeset_SecondScreen_trunk_Libraries_libcodegen_r51391.diff 836 B View Download
在 2013-06-27 22:06:47，"Andrew Nesbit"anesbit@echonest.com写道： - 显示引用文字 -