*.pdf files are for HTS and I didn't use them with
mlpg of SPTK . I wrote the Mean and varience values
from a traind HMM file in this order:
mean(static)[1](1),...,mean(static)[M](1),mean(delta)[1](1),...,mean(delta)[M](1),mean(delta^2)[1](1),...,mean(delta^2)[M](1),
var(static)[1](1),...,var(static)[M](1),var(delta)[1](1),...,var(delta)[M](1),var(delta^2)[1](1),...,var(delta^2)[M](1),
mean(static)[1](2),...,mean(static)[M](2),mean(delta)[1](2),...,mean(delta)[M](2),mean(delta^2)[1](2),...,mean(delta^2)[M](2),
var(static)[1](2),...,var(static)[M](2),var(delta)[1](2),...,var(delta)[M](2),var(delta^2)[1](2),...,var(delta^2)[M](2),
...
mean(static)[1](T),...,mean(static)[M](T),mean(delta)[1](T),...,mean(delta)[M](T),mean(delta^2)[1](T),...,mean(delta^2)[M](T),
var(static)[1](T),...,var(static)[M](T),var(delta)[1](T),...,var(delta)[M](T),var(delta^2)[1](T),...,var(delta^2)[M](T),
where M is the order of static feature vector,
mean(static)[m][t],
mean(delta)[m][t], and mean(delta^2)[m][t] means mean
value of static, delta and delta-delta of the m-th
order at t-th
frame, respectively.
as you told me . but the HMM file doesn't contain any
thing about frames 't' so I considered that t is the
state number and I repeat this format for t states
(instead of frame) with no headers , is that correct?