[hts-users:00625] 回复： Re: HInit Error

[Fang Li <lifang3cis@xxxxxxxxxxxx>]

Hi,Heiga
Thank for quick reply.
Yesterday HInit still can work,
today, I use the same data for HInit, including the labels and raw,
but it doesn't work 
The following is an example of mono label
3056679 4997647 a
it is longer than 5 frames.
the error message is as follows:
 ERROR [+2121]  HInit: Too Few Observation Sequences [0]
 FATAL ERROR - Terminating program /mnt/unzip/htk/HTKTools/bin/HInit
Error in /mnt/unzip/htk/HTKTools/bin/HInit  -A    -C /mnt/unzip/HTS/configs/trn.cnf -D -T 1 -S /mnt/unzip/HTS/data/scp/train.scp   -m 1 -u tmvw -w 3 -H /mnt/unzip/HTS/models/qst001/ver1/cmp/init.mmf -M /mnt/unzip/HTS/models /mnt/unzip/HTS/proto/ver1/state-5_stream-4_mcp-75_lf0-3.prtf -l a

best wishes
LiFang

"Heiga ZEN (Byung Ha CHUN)" <zen@xxxxxxxxxxxxxxx> 写道：
>  Hi,
>
> Fang Li wrote:
>
> > I also met the same error.
> > HInit can work well most of the time, but sometimes, it reports this 
> > error message.
> >
> > ERROR [+2121] HInit : Too Few Observation Sequences [0]
>
> This error means that there are no samples which can be used to train an HMM for this phoneme.
> If you are using the left-to-right 5-state HMM with no skip structure, 
> segments which are longer than 5 frames for this phoneme are used to train the HMM.
> If all segments are shorter than 5 frames, no segments can be used to estimate the HMM.
> So HInit returns the above error.
>
> Please check your database whether it contains any segments longer than 5 frames for this phoneme or not.
>
> Regards,
>
> Heiga ZEN (Byung Ha CHUN)
>
> -- 
> ------------------------------------------------
>  Heiga ZEN     (in Japanese pronunciation)
>  Byung Ha CHUN (in Korean pronunciation)
>
>  Department of Computer
>  Science and Engineering
>  Nagoya Institute of Technology
>  Gokiso-cho, Showa-ku, Nagoya 466-8555 Japan
>
>  http://www.sp.nitech.ac.jp/~zen
> ------------------------------------------------

---

雅虎免费邮箱-3.5G容量，20M附件