[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Although I am not familiar with IRCA1-noise I think you can generate this noise as follows.
You can analyse the spectra of a speech database (i.e. if you have the speech in a same vector x in Matlab you can do >X = fft(x)). And design an N-taps FIR filter. (>b = fir2(N,
2/length(X) :2/length(X):1,abs(X(1:length(X)/2)))) Then you can generate a random stationary noise (i.e. Matlab > n = randn(duration*samplerate,1). Finally filter the noise using the FIR (> y = filter(b,1,n)).
Of course it could be better if you use for example average spectra or 1/3 octave band spectra or even better using critical bands. You can implement them as in subtitle 2.2 in:
(It is only in Spanish but you it is not difficult to find bibliography on how to implement a DFT)
2012/5/2 ali fallah <ali.fallah@xxxxxxxxx>
I need a sample of " stationary noise with speech-like long-term spectrum / ICRA1-noise" . I have two wave files containing ICRA noise but the waveforms is not stationary.