Hi, Oura,
Thanks again for your reply. It's useful for me.
Regards
hlwang
At 2011-08-08 11:13:42,"Keiichiro Oura" <uratec@xxxxxxxxxxxxxxx> wrote:
>Hi,
>
>The main difference between mel-cepstrum from SPTK and STRAIGHT
>(Matlab) is the scale.
>In STRAIGHT, a type of normalization is implemented.
>But, I don't remember the details.
>
>Anyway, "hts_engine API" doesn't support aperiodic measure.
>Even if you solve the above problem, synthesized speech is not STRAIGHT voice.
>
>Regards,
>Keiichiro Oura
>
>
>2011/8/8 huanliangwang <huanliangwang@xxxxxxx>:
>> Hi, Oura:
>> "hts_engine API'' supports MCEP and f0 feature to synthesize speech.
>> But why dose it not support MCEP and f0 feature (excluding bap feature )
>> generated from STRAIGHT to synthesize speech? What is the main difference
>> between these features?
>>
>> Thanks
>>
>>
>> hlwang
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> At 2011-08-08 09:22:25,"Keiichiro Oura" <uratec@xxxxxxxxxxxxxxx> wrote:
>>>Hi,
>>>
>>>``hts_engine API'' doesn't support STRAIGHT.
>>>Please, use HMGenS and Matlab.
>>>
>>>Regards,
>>>Keiichiro Oura
>>>
>>>
>>>2011/8/8 huanliangwang <huanliangwang@xxxxxxx>:
>>>> Hi,
>>>> I also met with similar problem when using hts_engine to synthesize
>>>> wave in HTS-demo_CMU-ARCTIC-SLT_STRAIGHT.
>>>> Dose anyone know how to solve the problem or the reason it occurs?
>>>>
>>>> Regards,
>>>>
>>>> hlwang
>>>>
>>>> At 2011-08-08 01:49:09,"Kathrin Haag" <kathrin_haag@xxxxxxxxxxx> wrote:
>>>>
>>>> Hello,
>>>>
>>>> when I synthesise waveforms with hts_engine, I don't receive any errors, but
>>>> get waveforms without any signal as the output.
>>>> I used hts_engine for a different voice and got the correct output! I also
>>>> have all models I usually need for synthesis!
>>>>
>>>> I compared the wavefile sizes of the silent waveforms to the wavefile sizes
>>>> of my correctly synthesised waveforms from a different voice, and they all
>>>> have approximately the same size.
>>>>
>>>> Since I did not receive any substantial errors during training, I assume
>>>> that the error occurred during feature extraction at synthesis time. I
>>>> analysed the synthesised waveforms, and they don't have any feature values.
>>>>
>>>> I'd appreciate any suggestions on how to solve this problem.
>>>>
>>>> Thanks,
>>>>
>>>> Kathrin Haag
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>
>>
>>
>