Post by rich on Sept 5, 2008 5:30:07 GMT -5
I'm trying to implement an envelope measurement using 50% overlap-add FFT's. I zero the negative frequencies and reduce the DC bin by half then inverse transforming to get the quadrature pair in the time domain. This should create a Hilbert Transform pair in the real and imajinary halves of the inverse transformed data. The result I get shows some misbehavior in the time domain, though it's close to correct. The imaginary portion shows some beating at the block rate of the transform but the real does not. I'm windowing before the forward transform with a raised cosine window.
Admittedly I havent spent too many hours beating on it yet, but I'm wondering if there are any references on implementation that might save me some heartache and time. Everything I've seen covers the basic transform technique but skips the overlap-add case and any discussion of the effect of windowing.
Admittedly I havent spent too many hours beating on it yet, but I'm wondering if there are any references on implementation that might save me some heartache and time. Everything I've seen covers the basic transform technique but skips the overlap-add case and any discussion of the effect of windowing.