thank you a this work was a menu that done by yourself this of the
rich from these ridiculous but you couldn't travels or would be presenting
the goal here is basically to deal with a the problem of calibration or score
normalization and the process of a noisy speech
and
the framework the speaker pollution framework is i-vector a with cosine distance
so what we proposed to do is to estimate a noise i-vector using nonspeech portions
of the signal and use this to predict
the noise impacts on the i-vector space and on the score
so basically if we define x as the i-vector for clean speech with which is
not served and used i-vector from nonspeech we can estimate form
noisy portions and iced i-vector for noisy or observed speech
i'll five is directly off a standard deviation of the most signal in noisy speech
signal so it's related to snr so basically if for clean speech also is zero
and da actually the
the observed a an i-vector is actually equal to what we want to actually have
dark clean i-vector
and for extremely noisy is a speech we only observe that the noise what we
can try to due to the some a linear approximation
and use this to estimate it down the bias
a bias terminus scaling the term for a scoring function and use this vocal which
thank you