human voice will range from 300 - 3300 Hz, a filter first will eliminate
most of the irrelevant information.
then u can think of calculating the pitch of the signal,
if it is a human voice, the frequency spectrum will have peaks at the pitch
frequency and its harmonics.....
this may be a good approach...
Pradeep
On 6/25/07, Peter Gien wrote:
>
> I have an application that samples about 30 seconds of audio, 16 bit,
> 22050 samples per second. I would like to determine if there is a
> human voice signal present or not. With my limited knowledge, I would
> remove a DC offset on the entire signal, compute a 512 point FFT in a
> window that slides over the data. I would then check to see if the
> frequencies present correspond to those found in the human voice. This
> simplistic approach is probably not going to produce good
> discrimination.
>
> Does anyone know of a non-patented algorithm that can detect voice
> presence or lack thereof. The signal to noise ratio can range from
> very good to very poor.
>
> Thanks in advance.
>
--
If fate doesnt make u laugh, then u just dont get the joke.....