From: Mikolaj on
Dnia 03-11-2009 o 10:21:28 Ury <ujamshy(a)yahoo.com> napisa�(a):

>>>
>>>
>>> Ury wrote:
>>>
>>>> Hi,
>>>>
>>>> I receive a stream of linear 16 bit audio and need to identify the
>>>> possibly "interesting" parts vs. plain silence or noise.
>>>>
>>>> Just computing the average energy and comparing to a threshold is not
>>>> sifficient since there may be a background noise.
>>>>
>>>> Is there a lightweight adaptive energy detection algorithm?
>>>
>>> A very generic way is compare the long term average to the short term
>>> average.
>>
>> While that is the typical way of expressing the idea, I think "long
> term"
>> confuses the beginner. I think its more descriptive to say you need to
>> compare the medium term average with the short term average. In other
> words
>> look for surges in the energy above the gently rolling changes typical
> of
>> background noise.
>>
>> Steve
>>
>>
>
> What are typical durations for medium term and short term?
>
> Thanks,
> Ury.

1 and 10

--
Mikolaj