Hi What is the best format to record Speech files today for research work, MP3 or .wav or anything else? Thanks
What is the best format to record Speech files today for research work
Started by ●November 8, 2005
Reply by ●November 8, 20052005-11-08
2006 wrote:> Hi > > What is the best format to record Speech files today for research work, > MP3 or .wav or anything else? > > ThanksDepend on what you mean by "best" and what you mean by "research". MP3 takes less space, .wav is most accurate, but if you just want to listen to it, MP3 is more than fine. Mark
Reply by ●November 9, 20052005-11-09
Don't know your area, but I would still go ahead and give some suggestions. 16-bit linear PCM is probably the most popular format for working with speech algorithms. You can use it either in "raw" format or add a header like in "wav" format. For most work using wav files is a good idea. It interfaces very nicely with matlab too. Raw PCM is also a popular format for development with speech algorithms, especially if you are working in the same format/sampling rate etc. all the time. So it all depends on whether you find the header useful. Regards Piyush 2006 wrote:> Hi > > What is the best format to record Speech files today for research work, > MP3 or .wav or anything else? > > Thanks
Reply by ●November 9, 20052005-11-09
"Mark" <makolber@yahoo.com> wrote in message news:1131504524.738479.294250@g14g2000cwa.googlegroups.com...> > 2006 wrote: >> Hi >> >> What is the best format to record Speech files today for research work, >> MP3 or .wav or anything else? >> >> Thanks > > Depend on what you mean by "best" and what you mean by "research". > > > MP3 takes less space, .wav is most accurate, but if you just want to > listen to it, MP3 is more than fine.Unless you really need the smaller file sizes that you get from lossy compression, I would avoid it. You can always batch convert your wave files to MP3 down the road, but you can't go the other way without some losses (though they may be small). It is also much easier to read/decode a wave file if you need to worry about writing your own software.
Reply by ●November 10, 20052005-11-10
2006 wrote:> > Hi > > What is the best format to record Speech files today for research work, > MP3 or .wav or anything else?NIST WAV is very common (and not the same as microsoft WAV). It has a text header format which allows freeform text to be stored. This file format is quite well supported by libsndfile: http://www.mega-nerd.com/libsndfile/ Erik -- +-----------------------------------------------------------+ Erik de Castro Lopo +-----------------------------------------------------------+ GPLG GPLGPLGP GPLGPLGPLGP GPLGP GPL MICROSOFT GPLGP GPLGPLGPLGP GPLGPLGPL GPLGPL
Reply by ●November 10, 20052005-11-10
This is a multi-part message in MIME format. --------------000300000803080600020103 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Erik de Castro Lopo wrote:>2006 wrote: > > >>Hi >> >>What is the best format to record Speech files today for research work, >>MP3 or .wav or anything else? >> >> > >NIST WAV is very common (and not the same as microsoft WAV). > >It has a text header format which allows freeform text to >be stored. This file format is quite well supported by >libsndfile: > > http://www.mega-nerd.com/libsndfile/ > >Erik > >Quite true. Simply storing 16 bit data in big-endian order is also common. A lot of the ITU and TIA test data for speech codecs, VAD evaluation, etc. is simply stored like that. Steve --------------000300000803080600020103 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit <!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> <html> <head> <meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type"> </head> <body bgcolor="#ffffff" text="#000000"> Erik de Castro Lopo wrote: <blockquote cite="mid4372EA03.8852D291@mega-nerd.com" type="cite"> <pre wrap="">2006 wrote: </pre> <blockquote type="cite"> <pre wrap="">Hi What is the best format to record Speech files today for research work, MP3 or .wav or anything else? </pre> </blockquote> <pre wrap=""><!----> NIST WAV is very common (and not the same as microsoft WAV). It has a text header format which allows freeform text to be stored. This file format is quite well supported by libsndfile: <a class="moz-txt-link-freetext" href="http://www.mega-nerd.com/libsndfile/">http://www.mega-nerd.com/libsndfile/</a> Erik </pre> </blockquote> Quite true. Simply storing 16 bit data in big-endian order is also common. A lot of the ITU and TIA test data for speech codecs, VAD evaluation, etc. is simply stored like that.<br> <br> Steve<br> <br> </body> </html> --------------000300000803080600020103--
Reply by ●November 12, 20052005-11-12
2006 wrote:> Hi > > What is the best format to record Speech files today for research work, > MP3 or .wav or anything else? > > ThanksI don't know what you're trying to do, but i'd choose .wav or any other PCM format, because these formats are so simple that you can write yourself a function to read them and write them (well you can too for MP3 but it's more complicated, reading/writing a .wav file is a newbie thing)
Reply by ●November 13, 20052005-11-13
In article <1131504524.738479.294250@g14g2000cwa.googlegroups.com>, "Mark" <makolber@yahoo.com> wrote:>MP3 takes less space, .wav is most accurate, but if you just want to >listen to it, MP3 is more than fine.There is significant degradation in the conversion of wav to mp3. It is very easy to tell the difference between the two in simple listening tests even if you have old ears like mine.
Reply by ●November 13, 20052005-11-13
John Herman wrote:> In article <1131504524.738479.294250@g14g2000cwa.googlegroups.com>, "Mark" <makolber@yahoo.com> wrote:>>MP3 takes less space, .wav is most accurate, but if you just want to >>listen to it, MP3 is more than fine.> There is significant degradation in the conversion of wav to mp3. It is very > easy to tell the difference between the two in simple listening tests even if > you have old ears like mine.Well, there are different levels of MP3. If you use the lower compression modes you will have a much harder time telling the difference. -- glen
Reply by ●November 13, 20052005-11-13
glen herrmannsfeldt wrote:> John Herman wrote: > >> In article <1131504524.738479.294250@g14g2000cwa.googlegroups.com>, >> "Mark" <makolber@yahoo.com> wrote: > > >>> MP3 takes less space, .wav is most accurate, but if you just want to >>> listen to it, MP3 is more than fine. > > >> There is significant degradation in the conversion of wav to mp3. It >> is very easy to tell the difference between the two in simple >> listening tests even if you have old ears like mine. > > > Well, there are different levels of MP3. If you use the lower > compression modes you will have a much harder time telling the difference.Remember that MP3 perceptually compresses. What sounds good to us may not be that faithful to the original. When the goal is to capture data for research purposes, it makes no sense to use lossy compression and possibly skew the results. Steve
