[Subject Prev][Subject Next][Thread Prev][Thread Next][Date Index][Thread Index]

[hts-users:03738] Re: Are HMGenS -s states internally modified with -c 0 option?


Hi Matt

Thank you for the information. I'm sure I would have tripped on that format difference.

After browsing through the HMGenS code I began to worry that the state alignment was being used to independently specify the state duration sequence, but then subsequent processes could modify the actual state assignments. Given your comment, I am less concerned. Anyway, I should be able to discern this with some experimentation.

On a related note, do you know if the -s option can be used with -c 1 or -c 2 (in which case I expect it would be interpreted as an initial condition)?

Many thanks,
David
________________________________________
From: Matt Shannon [sms46@xxxxxxxxx]
Sent: Monday, April 22, 2013 2:49 AM
To: hts-users@xxxxxxxxxxxxxxx
Subject: [hts-users:03730] Re: Are HMGenS -s states internally modified with -c 0 option?

Hi David,

What is the distinction you are drawing between the actual state
sequence used for parameter generation and the state durations?  They
seem "equivalent" to me.

If you're just wondering about the format to provide the labels in, I
believe that when using '-s' you should use a label file in the format:

0 100000 state[2] pau
100000 150000 state[3]
150000 200000 state[4]
200000 1450000 state[5]
1450000 1550000 state[6]
1550000 1600000 state[2] f
1600000 1650000 state[3]
1650000 1850000 state[4]
1850000 1950000 state[5]
1950000 2150000 state[6]

instead of in the format:

0 1550000 pau
1550000 2150000 f

Thanks,

Matt


On 19/04/13 19:15, David Klein wrote:
> Hi
>
> I would like to try to use HMGenS to run MLPG with a state sequence that is given by an external application (presuming I use -c 0 option).
>
> I think I can do this through the -s option. However, it is not completely clear to me if this is used to specify the actual state sequence to use for parameter generation (which I would like), or if it merely is used to specify the state durations.
>
> Is -s the correct way to specify a given state sequence to be used for parameter generation, which won't be internally modified?
>
> Thanks
> David
>


References
[hts-users:03725] Are HMGenS -s states internally modified with -c 0 option?, David Klein
[hts-users:03730] Re: Are HMGenS -s states internally modified with -c 0 option?, Matt Shannon