that's is the improvements on the bottleneck features itself work by some the end from
us to use a couple of the search for a million myself and dinally one
and if you follow us onions work over the past few years to say these
make great gains in using deep bottleneck features for lid
so this particular paper
it extends from some work that is published i think last year just pretty much
this work and he's using the bottleneck features vector and bottleneck layer and fifty features
to create to extract i-vectors what he's doing this is basically
taking out that p gmm of putting in a phonetic mixture french analyses in its
place and what this does is it allows the
single step to do the analysis feature reduction
and a combination they're also on the locks some efficiency gains that allows them to
explore and doing something like sdc with take bottleneck features that is concatenating or extending
the context
time which appears to not quite well
the test is done on lre zero nine with the six most highly confused languages
and he's got some improvement gains and as you'll see if you come to the
poster the improvement is less
alpha three seconds and it is for the longer utterances that's not really surprising but
if you're interested where poster number eleven