Open tomfrancart opened 6 years ago
I think this would approximately boil down to: -adapting the speech for negative SNRs, as the total RMS is then mostly determined by the noise (at the limit for SNR = 0, the speech only adds+3dB) -adapting the noise for positive SNRs, as the total RMS is then mostly determined by the speech. So might not be such a useful feature (?)
Good point. However, the goal is to have a procedure that can be used in the clinic at any SNR, so perhaps we should implement it as you suggest... I wonder how this is done for DTT, @mh21 or @Gundersanne ?
For speech in noise experiments, it's sometimes desired to adapt both the speech and the noise level, in order to keep a constant total RMS. Either add this to the adaptive procedure, or add a plugin procedure to do this.