AAC | ADPCM | Convolution | DAFx | FFT | IIR | Mixer | MP3 | MPEG | MPEG-4

Hi,
Please, can anybody tell how can I remove the hiss noise from the interpolated audio signal?
Here, I am simulating the interpolation using c-program and scilab code. I Though, I have used
scilab inbuild cubic interpolation function, noise is there in the output audio signal.
Also, I am using low pass filter after interpolation. Since I am interpolating 8KHz sample rate
audio signal to 32KHz, I kept cutoff of the lowpass 4KHz.
Is there any mistake in my design? Please let me know.
Thanks
	

(You need to be a member of audiodsp -- send a blank email to audiodsp-subscribe@yahoogroups.com )

Hello sunil@suni...

> Please, can anybody tell how can I remove the hiss noise from the
> interpolated audio signal? Here, I am simulating the interpolation using
> c-program and scilab code. I Though, I have used scilab inbuild cubic
> interpolation function, noise is there in the output audio signal. Also, I
> am using low pass filter after interpolation. Since I am interpolating
> 8KHz sample rate audio signal to 32KHz, I kept cutoff of the lowpass 4KHz.
> Is there any mistake in my design?

For integer ratio interpolation the best approach (IMHO) is to use filter
based interpolation. You have interpolation with rate 4, so you have to
insert three 0 samples after every input signal sample. Then you pass
resulting signal through lowpass filter with cutoff at 4 kHz. This technique
is discussed for example in "Understanding Digital Signal Processing" by R.
G. Lyons. Note that if the bandwidth of your signal ends sharp at 4 kHz, you
need very sharp filter as well. In usual telecom signals bandwidth ends at
3,5 kHz, so you have 1 kHz for the filter transition band, and no aliasing
occurs. If 3,5 kHz is not your case you can:

1) Use extremely long FIR filter at a cost of computations required.
2) Pre-filter input signal with 3,5 kHz lowpass filter at a cost of
bandwidth.
3) Try to use high order IIR filter at a cost of phase distortions.
4) Use Frequency Response Masking filter, which is similar to 1), but
computational costs are significantly reduced.

In my opinion using polynomial interpolation for fixed (and rational) rate audio 
resampling is not economical. It can be useful for not fixed and/or not
rational resampling, but high quality requires oversampling (zero samples
insertion) first and then high-order polynomial. It is deeply discussed in

http://www.biochem.oulu.fi/~oniemita/dsp/deip.pdf

Regards
-- 
Grzegorz "Krashan" Kraszewski
krashan@kras...
	

(You need to be a member of audiodsp -- send a blank email to audiodsp-subscribe@yahoogroups.com )