Oh... I thought it was derived from Maia's ML excitation model. Is it to some extent relevant?
I would check MELP spec. further.
Na Xingyu
At 2011-07-21 15:09:21,"Keiichiro Oura" <uratec@xxxxxxxxxxxxxxx> wrote:
>Hi,
>
>lpf is not statistical model.
>So, you cannot train it.
>
>Analog-to-Digital Conversion of Voice by 2,400 bit/second Mixed
>Excitation Linear Prediction (MELP). United States Department of
>Defense. Dec. 1999.
>
>Regards,
>Keiichiro Oura
>
>
>2011/7/19 那兴宇 <nxy-yzqs@xxxxxxx>:
>> Hi,
>>
>> In the new demo and engine, lpf model is supported. However, it is generated
>> by the makefilter perl script. How to train a CD decision tree based lpf
>> model?
>> Please give a hint. Thanks
>>
>> Na Xingyu
>>
>>
>