DSPRelated.com
Forums

Clean speech wav files

Started by doggie April 3, 2006
Hi everyone,do any of you know where i can get clean speech wav files that
are more than 10 seconds? I need them for testing my speech detection
algorithms. Thanks


Why not record some off of talk radio or get an audio book?

-Clark

"doggie" <elusivetruelove2003@yahoo.com> wrote in message
news:qpudnUW_5usQn6zZRVn-rQ@giganews.com...
> > Hi everyone,do any of you know where i can get clean speech wav files that > are more than 10 seconds? I need them for testing my speech detection > algorithms. Thanks > >
http://www.phon.ucl.ac.uk/resource/data.html

>http://www.phon.ucl.ac.uk/resource/data.html > >
nice site..but its a pity they are only for sale. I'm actually looking for free clean speech wave file. Thanks anyway :)
doggie wrote:
> Hi everyone,do any of you know where i can get clean speech wav files that > are more than 10 seconds? I need them for testing my speech detection > algorithms. Thanks
Hook a microphone to your sound card and infiltrate a cocktail party. Jerry -- Engineering is the art of making what you want from things you can get. &#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;
Jerry Avins wrote:
> doggie wrote: > >> Hi everyone,do any of you know where i can get clean speech wav files >> that >> are more than 10 seconds? I need them for testing my speech detection >> algorithms. Thanks > > > Hook a microphone to your sound card and infiltrate a cocktail party. > > Jerry
I've just recently been transferring cassette material to .mp3, so any comedy (without laugh-track) or drama items might do. I've been recording to .wav as a first step, followed by conversion to .mp3. Also, while I was experimenting initially, I seem to remember that the Windows built-in "Recorder" function (under audio support stuff) will do up to 60 seconds....except I don't remember its default format.... Geoff.
Thanks everyone for your help. I've found another way.i'll just record it
on my mobile phone and transfer to my computer. :)

By the way,i have a few other simple questions that i hope you will give
me your guidance.

1)A lot of algorithms mentioned preprocessing. However,sometimes it is a
low pass filtering and sometimes it is a high pass filtering. I feel that
it should be a bandpass filter that filters the noise outside the range
of
human speech. If not, then what is preprocessing used for? Please advice.

2)i am trying to add noise using matlab but i do not really understand the
Help file. How should i add gaussian noise of a certain db? i tried using
wgn(length(cleanspeech),1,0.001). But it completely cover my speech. Isn't
0.001 db not low enough? Please advise.

Thanks a lot.
"doggie" <elusivetruelove2003@yahoo.com> writes:
> [...] > 2)i am trying to add noise using matlab but i do not really understand the > Help file. How should i add gaussian noise of a certain db? i tried using > wgn(length(cleanspeech),1,0.001). But it completely cover my speech. Isn't > 0.001 db not low enough? Please advise.
According to the Mathworks help for wgn(m,n,p), "p specifies the power of y in decibels relative to a watt" Power p in dB (decibels) relative to one watt is related to linear power p_l as follows: p = 10 * log_10(p_l) If we solve for p_l: p_l = 10^(p/10). So we have the following conditions: p > 0 ==> p_l > 1 watt p = 0 ==> p_l = 1 watt p < 0 ==> p_l < 1 watt In words, you have to use negative numbers to get small values of linear power. The more negative, the smaller the linear power. -- % Randy Yates % "Watching all the days go by... %% Fuquay-Varina, NC % Who are you and who am I?" %%% 919-577-9882 % 'Mission (A World Record)', %%%% <yates@ieee.org> % *A New World Record*, ELO http://home.earthlink.net/~yatescr
>"doggie" <elusivetruelove2003@yahoo.com> writes: >> [...] >> 2)i am trying to add noise using matlab but i do not really understand
the
>> Help file. How should i add gaussian noise of a certain db? i tried
using
>> wgn(length(cleanspeech),1,0.001). But it completely cover my speech.
Isn't
>> 0.001 db not low enough? Please advise. > >According to the Mathworks help for wgn(m,n,p), > > "p specifies the power of y in decibels relative to a watt" > >Power p in dB (decibels) relative to one watt is related to >linear power p_l as follows: > > p = 10 * log_10(p_l) > >If we solve for p_l: > > p_l = 10^(p/10). > >So we have the following conditions: > > p > 0 ==> p_l > 1 watt > p = 0 ==> p_l = 1 watt > p < 0 ==> p_l < 1 watt > >In words, you have to use negative numbers to get small values of >linear power. The more negative, the smaller the linear power. >-- >% Randy Yates % "Watching all the days go by... >%% Fuquay-Varina, NC % Who are you and who am I?" >%%% 919-577-9882 % 'Mission (A World Record)', >%%%% <yates@ieee.org> % *A New World Record*, ELO >http://home.earthlink.net/~yatescr >
ok.let's say if my signal power is 10 db. and i want my noise to 5db to get an SNR of 5db. Do i use wgn(length(cleanspeech),1,5) then? Thanks.
"doggie" <elusivetruelove2003@yahoo.com> writes:

>>"doggie" <elusivetruelove2003@yahoo.com> writes: >>> [...] >>> 2)i am trying to add noise using matlab but i do not really understand > the >>> Help file. How should i add gaussian noise of a certain db? i tried > using >>> wgn(length(cleanspeech),1,0.001). But it completely cover my speech. > Isn't >>> 0.001 db not low enough? Please advise. >> >>According to the Mathworks help for wgn(m,n,p), >> >> "p specifies the power of y in decibels relative to a watt" >> >>Power p in dB (decibels) relative to one watt is related to >>linear power p_l as follows: >> >> p = 10 * log_10(p_l) >> >>If we solve for p_l: >> >> p_l = 10^(p/10). >> >>So we have the following conditions: >> >> p > 0 ==> p_l > 1 watt >> p = 0 ==> p_l = 1 watt >> p < 0 ==> p_l < 1 watt >> >>In words, you have to use negative numbers to get small values of >>linear power. The more negative, the smaller the linear power. >>-- >>% Randy Yates % "Watching all the days go by... >>%% Fuquay-Varina, NC % Who are you and who am I?" >>%%% 919-577-9882 % 'Mission (A World Record)', >>%%%% <yates@ieee.org> % *A New World Record*, ELO >>http://home.earthlink.net/~yatescr >> > > ok.let's say if my signal power is 10 db. and i want my noise to 5db to > get an SNR of 5db. Do i use wgn(length(cleanspeech),1,5) then? Thanks.
Yes. But make sure you compute your signal power the same way Matlab is computing the wgn power: Psignal = 10*log10(sum(x.*x)/length(x)); -- % Randy Yates % "How's life on earth? %% Fuquay-Varina, NC % ... What is it worth?" %%% 919-577-9882 % 'Mission (A World Record)', %%%% <yates@ieee.org> % *A New World Record*, ELO http://home.earthlink.net/~yatescr