[Tftb-help] RE: Tftb-help Digest, Vol 4, Issue 3

[Eric Chassande-Mottin]

> >this is one solution. another simpler one is to keep the long signal as
> >it is, to precise at which time instants the spectrogram should be
> >computed and with which frequency resolution.

> Thanks for your reply. but I cannot find how help at "help tfrsp" for my
> question. Is there any particular reason to choose the window length=
> signal length/4 ?

here are the first few lines of the online help of tfrsp

TFRSP   Spectrogram time-frequency distribution.
        [TFR,T,F]=TFRSP(X,T,N,H,TRACE) computes the Spectrogram
        distribution of a discrete-time signal X.

        X     : signal.
        T     : time instant(s)          (default : 1:length(X)).
        N     : number of frequency bins (default : length(X)).

my comment was referring to the last two.

you can set T and N so that the output TF matrix is of reasonable size
while giving a sufficient resolution for your needs.

see the following example

M=10000;
x=fmlin(M);
t=1:100:M; N=512;
tfrsp(x,t,N);

the spectrogram is computed every 100 samples of the input signal (as
defined by t) and N=512 frequency bins. therefore, the output matrix is of
size M/100 times N which is much small than M^2. obviously the resolution
is smaller, but visually the result is sufficient to make some analysis.

eric.