audacity-nyquist

 [Audacity-nyquist] Statistical analysis (and hello) From: George Jenner - 2007-01-11 06:28:12 Attachments: Message as HTML ```Hello, I was just searching the archives to see if I could do some statistical analysis of recordings. I see Edgar recently published an A-weighting filter and I've succesfully runi it as my first plug in. (Sorry I've only just joined the list so I only know you by the archives). At work I'm using audacity to measure noise levels, so I wonder if this has been done before. I'd like to determine L10, L90 and Leq of a sample. For those of you whose jobs are more interesting than mine, L10 is the level which is exceeded by 10% of all samples. Has this been done by anyone? Lisp is a foreign language to me so I'll need some time to get used to it. Thanks for any help, George ```
 Re: [Audacity-nyquist] Statistical analysis (and hello) From: Alex S. Brown, PMP - 2007-01-12 04:28:36 ```I did a silence-detection plug-in which is now included in the regular Audacity distribution. That required me to look at some very basic noise detection issues. I did not have to do the statistical analysis, because I just looked at areas of the signal that exceeded or fell below a set threshold. You could look at my plug-in for help. I found one of the hardest parts of writing it was being careful not to accumulate all the samples in memory. If you do, then your memory use quickly skyrockets for anything more than a very short sound wave. Your first step should be to look at the built-in functions to manipulate the signal. If you can find a way to reduce the number of samples you have to read, you will get a result much more quickly. If you need specific advice, let us know more about the math of your calculation and how it is typically done, and we might be able to guide you towards some useful built-in functions or programming models. --Alex George Jenner wrote: > Hello, > > I was just searching the archives to see if I could do some > statistical analysis of recordings. I see Edgar recently published an > A-weighting filter and I've succesfully runi it as my first plug in. > (Sorry I've only just joined the list so I only know you by the > archives). > > At work I'm using audacity to measure noise levels, so I wonder if > this has been done before. I'd like to determine L10, L90 and Leq of > a sample. For those of you whose jobs are more interesting than mine, > L10 is the level which is exceeded by 10% of all samples. > > Has this been done by anyone? Lisp is a foreign language to me so > I'll need some time to get used to it. > > Thanks for any help, > George -- Alex S. Brown, PMP alexsbrown@... http://www.alexsbrown.com - Free PM Articles http://www.rlprj.com - PM Speaking and Teaching ahttps://www.xing.com/profile/AlexS_Brown OpenBC/XING Professional Networking http://www.linkedin.com/in/alexsbrown LinkedIn Networking ```
 Re: [Audacity-nyquist] Statistical analysis (and hello) From: Mikael - 2007-01-12 13:19:03 Attachments: Message as HTML ```Hi George! I have fiddled a bit with acoustics in audacity with some success and can help out with the equivalent level. You can use the nyquist function snd-avg to get it like this: (setq sl (snd-length s 999999999999)) (setq calibration 97) (+ calibration (* 0.5 (linear-to-db (snd-maxsamp (snd-avg (mult s s) sl sl OP-AVERAGE) )))) Paste this into the Effect -> Nyquist prompt, and you'll get the equivalent level in dB of what ever mono audio you selected. Of course it is uncalibrated, that's why I have the calibration variable if you would like to set it. Using 97 like I do seems to give 94 dB for a 1000 Hz tone with amplitude 1.0. Note that it is the equivalent level without A-weighting, to get that you need to filter with an A-filter first (se the post by Edgar you mention). Watch out for clipping when filtering, I believe audacity clips any sample above 1.0... I would guess that L10 and L90 are more diffcult, I think you need to calculate the histogram to get it. Please mail any success to this list. Mikael Ogren PS I use snd-maxsamp to extract the single sample from the previous step, I guess there is a more elegant way but I couldn't find it. DS > > > On 1/12/07, Alex S. Brown, PMP wrote: > > > > I did a silence-detection plug-in which is now included in the regular > > Audacity distribution. That required me to look at some very basic noise > > detection issues. I did not have to do the statistical analysis, because > > > > I just looked at areas of the signal that exceeded or fell below a set > > threshold. > > > > You could look at my plug-in for help. I found one of the hardest parts > > of writing it was being careful not to accumulate all the samples in > > memory. If you do, then your memory use quickly skyrockets for anything > > more than a very short sound wave. > > > > Your first step should be to look at the built-in functions to > > manipulate the signal. If you can find a way to reduce the number of > > samples you have to read, you will get a result much more quickly. > > > > If you need specific advice, let us know more about the math of your > > calculation and how it is typically done, and we might be able to guide > > you towards some useful built-in functions or programming models. > > > > --Alex > > > > George Jenner wrote: > > > Hello, > > > > > > I was just searching the archives to see if I could do some > > > statistical analysis of recordings. I see Edgar recently published an > > > > > A-weighting filter and I've succesfully runi it as my first plug in. > > > (Sorry I've only just joined the list so I only know you by the > > > archives). > > > > > > At work I'm using audacity to measure noise levels, so I wonder if > > > this has been done before. I'd like to determine L10, L90 and Leq of > > > a sample. For those of you whose jobs are more interesting than mine, > > > L10 is the level which is exceeded by 10% of all samples. > > > > > > Has this been done by anyone? Lisp is a foreign language to me so > > > I'll need some time to get used to it. > > > > > > Thanks for any help, > > > George > > > > -- > > Alex S. Brown, PMP > > alexsbrown@... > > http://www.alexsbrown.com - Free PM Articles > > http://www.rlprj.com - PM Speaking and Teaching > > ahttps://www.xing.com/profile/AlexS_Brown OpenBC/XING Professional > > Networking > > http://www.linkedin.com/in/alexsbrown LinkedIn Networking > > > > > > ------------------------------------------------------------------------- > > > > Take Surveys. Earn Cash. Influence the Future of IT > > Join SourceForge.net's Techsay panel and you'll get the chance to share > > your > > opinions on IT & business topics through brief surveys - and earn cash > > > > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV > > _______________________________________________ > > Audacity-nyquist mailing list > > Audacity-nyquist@... > > https://lists.sourceforge.net/lists/listinfo/audacity-nyquist > > > > ```
 Re: [Audacity-nyquist] Statistical analysis (and hello) From: George Jenner - 2007-01-13 13:23:12 Attachments: Message as HTML ```Thanks to you both for replying. Alex I've seen your silence plugin and it does look like a good place to start. We tend to work with 1/8second samples for a sound level, all averaged over 15 minutes, so what I think I'll do is this: read 1/8 second sample; average the 1/8 second (thanks Mikael, will look at your solution) - that way I don't have to worry about memory by analysing for 15 minutes; store the level in an array (I presume this is just a list in lisp); repeat for next 1/8 seconds until I get to 15 minutes; sort the array from lowest to highest; find the percentiles I'm interested in; average them all for an overall 15minute average; output the results, preferable to a text file. I'll probably have multiple files to do - 15 minute samples over at least a day, so I'd have to load the samples, analyse, write the stats, dump the sample etc. So tomorrow I learn lisp (or the next day). Thanks again George On 1/13/07, Mikael wrote: > > Hi George! > > I have fiddled a bit with acoustics in audacity with some success and can > help out with the equivalent level. You can use the nyquist function snd-avg > to get it like this: > > (setq sl (snd-length s 999999999999)) > (setq calibration 97) > (+ calibration (* 0.5 (linear-to-db (snd-maxsamp (snd-avg (mult s s) sl sl > OP-AVERAGE) )))) > > Paste this into the Effect -> Nyquist prompt, and you'll get the > equivalent level in dB of what ever mono audio you selected. Of course it is > uncalibrated, that's why I have the calibration variable if you would like > to set it. Using 97 like I do seems to give 94 dB for a 1000 Hz tone with > amplitude 1.0. > > Note that it is the equivalent level without A-weighting, to get that you > need to filter with an A-filter first (se the post by Edgar you mention). > Watch out for clipping when filtering, I believe audacity clips any sample > above 1.0... > > I would guess that L10 and L90 are more diffcult, I think you need to > calculate the histogram to get it. Please mail any success to this list. > > Mikael Ogren > > PS I use snd-maxsamp to extract the single sample from the previous step, > I guess there is a more elegant way but I couldn't find it. DS > > > > > > > On 1/12/07, Alex S. Brown, PMP wrote: > > > > > > I did a silence-detection plug-in which is now included in the regular > > > Audacity distribution. That required me to look at some very basic > > > noise > > > detection issues. I did not have to do the statistical analysis, > > > because > > > I just looked at areas of the signal that exceeded or fell below a set > > > threshold. > > > > > > You could look at my plug-in for help. I found one of the hardest > > > parts > > > of writing it was being careful not to accumulate all the samples in > > > memory. If you do, then your memory use quickly skyrockets for > > > anything > > > more than a very short sound wave. > > > > > > Your first step should be to look at the built-in functions to > > > manipulate the signal. If you can find a way to reduce the number of > > > samples you have to read, you will get a result much more quickly. > > > > > > If you need specific advice, let us know more about the math of your > > > calculation and how it is typically done, and we might be able to > > > guide > > > you towards some useful built-in functions or programming models. > > > > > > --Alex > > > > > > George Jenner wrote: > > > > Hello, > > > > > > > > I was just searching the archives to see if I could do some > > > > statistical analysis of recordings. I see Edgar recently published > > > an > > > > A-weighting filter and I've succesfully runi it as my first plug in. > > > > (Sorry I've only just joined the list so I only know you by the > > > > archives). > > > > > > > > At work I'm using audacity to measure noise levels, so I wonder if > > > > this has been done before. I'd like to determine L10, L90 and Leq > > > of > > > > a sample. For those of you whose jobs are more interesting than > > > mine, > > > > L10 is the level which is exceeded by 10% of all samples. > > > > > > > > Has this been done by anyone? Lisp is a foreign language to me so > > > > I'll need some time to get used to it. > > > > > > > > Thanks for any help, > > > > George > > > > > > -- > > > Alex S. Brown, PMP > > > alexsbrown@... > > > http://www.alexsbrown.com - Free PM Articles > > > http://www.rlprj.com - PM Speaking and Teaching > > > ahttps://www.xing.com/profile/AlexS_Brown OpenBC/XING Professional > > > Networking > > > http://www.linkedin.com/in/alexsbrown LinkedIn Networking > > > > > > > > > ------------------------------------------------------------------------- > > > > > > Take Surveys. Earn Cash. Influence the Future of IT > > > Join SourceForge.net's Techsay panel and you'll get the chance to > > > share your > > > opinions on IT & business topics through brief surveys - and earn cash > > > > > > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV > > > > > > _______________________________________________ > > > Audacity-nyquist mailing list > > > Audacity-nyquist@... > > > https://lists.sourceforge.net/lists/listinfo/audacity-nyquist > > > > > > > > > ------------------------------------------------------------------------- > Take Surveys. Earn Cash. Influence the Future of IT > Join SourceForge.net's Techsay panel and you'll get the chance to share > your > opinions on IT & business topics through brief surveys - and earn cash > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV > > _______________________________________________ > Audacity-nyquist mailing list > Audacity-nyquist@... > https://lists.sourceforge.net/lists/listinfo/audacity-nyquist > > > ```