I have two doubts:
1. Why are only 512 energy points (calculated from 1024 point fft) used in the threshold calculation and not 1024?
2.The partition tables in the model do not divide 512 frequencies into partitions. instead it divides about 480 frequency lines.
In effect the sum of the the total lines in 63(for sampling frequency -44100) partitions is not equal to 512.
Full Version: threshold calculation in physchoacoustic model 2
QUOTE(abhijit @ Jun 15 2007, 10:47) 
1. Why are only 512 energy points (calculated from 1024 point fft) used in the threshold calculation and not 1024?
That's just how the FFT works. You don't need the other half because it doesn't contain any more useful informations due to your signal being real-valued (in contrast to complex valued). The upper part is just a mirrored version of the lower part. bin 0-512 already span all the frequencies of interest.
QUOTE(abhijit @ Jun 15 2007, 10:47)
2.The partition tables in the model do not divide 512 frequencies into partitions. instead it divides about 480 frequency lines.
I don't think that this is a big problem. 480 is close enough to 512. It translates to a bandwidth of 20 kHz for a samplingrate of 44100 Hz. I can hardly hear anything above 18 kHz
Cheers!
SG
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.