Re: [linux-audio-dev] Resampling

New Message Reply About this list Date view Thread view Subject view Author view Other groups

Subject: Re: [linux-audio-dev] Resampling
From: Juhana Sadeharju (kouhia_AT_nic.funet.fi)
Date: Thu Aug 02 2001 - 23:02:24 EEST


>From: Richard Dobson <RWD_AT_cableinet.co.uk>
>
>I suspect there are other aspects too where the implementation couyld be
>improved. More orthodox pvocs, sucha s the F.R Moore one, and CARL (on
>which my 'PVStream' real-time pvoc is based), does offer exact
>reconstruction. Because of the interpolation, Csound pvoc probably never
>gives an exact reconstruction.

I will check those packages. I checked phase vocoder from a paper
written by Gordon and Strawn. It seems to remove phases entirely
and uses sin() generators in synthesis. Too large hops between
analysis windows could be another reason for bad quality.

I'm going to library and check Portnoff's original papers. There is
also one paper which discusses how window function types affect
the system.

Portnoff's analysis uses large Kaiser window, but after that something
weird happens: the M*N length windowed signal is time-aliased to
buffer of length N. All M subsignals are overlapped and added together.
I wondering if such operation smears the high-frequency analysis
because what really is analysed is a long period of the original signal.

Does it make difference if original signal is splitted to two half
bands (0-11050 Hz and 11050-22050 Hz) and then STFT analysis with
K length window is applied to lower band and K/2 length window analysis
is applied to the higher band? Somehow I feel the one-STFT process is not
doing the time-frequency resolution tradeoff properly, and logarithmic
splitting could help -- specially in case where signal modifications
are changing heavily.

Regards,

Juhana

74f6bb7f8ba0fecc27ab0ff61ed6367d


New Message Reply About this list Date view Thread view Subject view Author view Other groups

This archive was generated by hypermail 2b28 : Thu Aug 02 2001 - 23:04:06 EEST