In this paper, in the Formula at the beginning of 2.2, we have



$i=0,1$ - the number of probability formula

$O_t$ - the state at moment $t$

$b_i(O_t)$ - two probabilities or estimations for the state $O_t$

The result ($B$) is named "likelihood". How can likelihood be obtained from 2 numbers? Is this a weighted average like

$b_i(O_t)= 0*b_0(O_t) + 1*b_1(O_t)$


$b_i(O_t)= -1*b_0(O_t) + 1*b_1(O_t)$

or something?

  • $\begingroup$ It looks like some sort of indexing similar to the idea of vectors. $\endgroup$ – azetina Jul 5 '12 at 14:41

This is a good example of why it makes sense to quote texts with more context, since it's often the context that supplies the clues for interpretation.

$B$ is referred to not just as a likelihood, but as "the likelihood [...] of [...] the frame [...] being a speech or a noise frame". This is rather badly phrased (like some other things in that paper), but it's clear that a single number cannot be simultaneously the likelihood of being a speech or a noise frame, so $B$ is the pair of likelihoods for these two cases. The curly brace notation for this pair is somewhat unfortunate in my view.


My guess is that he meant it as $B_i:=b_i(O_t)$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.