Wikipedia:Reference desk/Archives/Miscellaneous/2022 December 20
Miscellaneous desk | ||
---|---|---|
< December 19 | << Nov | December | Jan >> | Current desk > |
Welcome to the Wikipedia Miscellaneous Reference Desk Archives |
---|
The page you are currently viewing is a transcluded archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages. |
December 20
[edit]Constant/Variable-Q transform with non-logarithmic frequency spacing
[edit]Most CQT implementations are done in a logarithmic frequency scale. However, I found out the algorithm is actually flexible (which was proven by spacing the frequency bins in Mel scale and setting window size for each Goertzel bands appropriately to provide equal frequency resolution on the aforementioned scale). 125.164.18.14 (talk) 12:05, 20 December 2022 (UTC)
- What is your question? AndyTheGrump (talk) 12:08, 20 December 2022 (UTC)
- @AndyTheGrump Is variable-Q transform with Mel scale better than typical Mel spectrogram (which although it uses triangular filter banks, this does not actually improve the temporal resolution on higher frequencies because it is calculated by summing FFT magnitude with filter banks, unlike Brown-Puckette method)? 2001:448A:3047:13E4:11FC:BB2B:C9B3:F4EE (talk) 10:31, 23 December 2022 (UTC) (actually, my IP has changed from my previous post)
- Mel scale is an empirical logarithmic law intended to plot the ability of an average human to distinguish changes in pitch. The Brown-Puckette algorithm (see abstract here) does employ an FFT followed by selective weighting of the output bins; this method is the one described here. An audio spectrogram may also be implemented with a set of individually tailored filters but using an FFT greatly simplifies computation. I won't declare any particular implementation as "best" in this field of psychoacoustics where skilled musicians have their opinions but I have these conclusions from my work with objective rating of telephone speech impairments:
- @AndyTheGrump Is variable-Q transform with Mel scale better than typical Mel spectrogram (which although it uses triangular filter banks, this does not actually improve the temporal resolution on higher frequencies because it is calculated by summing FFT magnitude with filter banks, unlike Brown-Puckette method)? 2001:448A:3047:13E4:11FC:BB2B:C9B3:F4EE (talk) 10:31, 23 December 2022 (UTC) (actually, my IP has changed from my previous post)
- There is not a universal exact agreement on a Mel scale
- My work employs the Bark scale that is similar to Mel. A simple conjecture is that the Mel scale works well for identifying pitch in musical (desired) sounds while the Bark scale works well at quantifying the annoyances caused by interfering sounds.
- I use a single FFT that is configured for real-only (no quadrature) input samples and sampling duration 100ms. This gives fractional-second temporal resolution and adequate frequency resolution (small bin separation) to handle the lowest audible frequencies. This is much finer frequency resolution than needed at higher frequencies.
- The phase information in each FFT output bin is discarded by squaring the complex number result.
- To distribute power on the required (Mel or Bark) scale, filters collect groups of FFT bins and add powers. The filters should overlap at their half-power ( -3dB ) points, proven by a linear freqency sweep test. Philvoids (talk) 17:03, 24 December 2022 (UTC)
- @Philvoids: There is a difference between summing the groups of FFT bins within a filter and using real/imag parts of FFT bins to calculate CQT/VQT, as described in [1]. The first one does not improve temporal resolution, while the latter actually does improve time resolution (or in other words, temporal resolution isn't being blurred in higher frequencies), which becomes obvious with a logarithmic frequency scale and FFT size of 44100 samples (1 sec at 44100 Hz sample rate) and higher. 2001:448A:3047:13E4:747F:CB26:3974:7FB1 (talk) 22:16, 26 December 2022 (UTC)
- I worked with voice telephony so my sampling was 100 msec at 8000 Hz sample rate, using only the 300 - 3200 Hz frequency range. Averaging 10 seconds telephone speech = total 100 FFTs (minus some synchronization signals) gives me in real time a single objective measure of impairments during a call i.e. any deviations from a stored 10-second reference speech. Philvoids (talk) 00:04, 27 December 2022 (UTC)
Polk HS and Andrew Johnson HS
[edit]In the Television show "Married with Children", the main character "Al Bundy" had as his shining moment scoring four touchdowns for Polk High School in the City Championship game against Andrew Johnson High School in 1966. It got me to thinking, In real life James Polk and Andrew Johnson are among the two US Presidents that I would expect to have the fewest High Schools named after them (and maybe Buchanan and WHarrison). If there *is* a city with both High Schools, I would expect it would be in Tennessee. So, any ideas on a better way than google to find a) Any cities that actually have a Polk High School (In google, I'm only finding a few and these are intermixed with "Direction" Polk High School from Polk County, Iowa) and b) whether there are any cities with high schools named after both? Naraht (talk) 14:15, 20 December 2022 (UTC)
- Courtesy link: Married... with Children.
- Whether there actually are or not (I will leave that to researchers on your side of the Pond), I suspect the show's writers would have wanted to pick a nonexistent combination, to avoid possible legal problems, non-legal controversy, or just distracting attention to the use of supposedly identifiable schools. Of course, 35+ years ago it might not have been easy to make a definitive search, so perhaps they (as you imply) just picked the least likely combination they could think of. {The poster formerly known as 87.81.230.195} 51.194.245.235 (talk) 15:59, 20 December 2022 (UTC)
- I can't help but be reminded of the grandiosely named George Bush Intercontinental Airport. Sagittarian Milky Way (talk) 20:03, 20 December 2022 (UTC)
- Here is the other useful information about April 27. --Askedonty (talk) 21:07, 20 December 2022 (UTC)
- I can't help but be reminded of the grandiosely named George Bush Intercontinental Airport. Sagittarian Milky Way (talk) 20:03, 20 December 2022 (UTC)
- Fictional locations are not unusual. The Law & Order franchise has often referred to the non-existent "Hudson University". Going farther back, there was the famous "Train leaving on Track 5 for Anaheim, Azusa and Cuc...amonga," Those are real cities, but there was no single train connecting all three. ←Baseball Bugs What's up, Doc? carrots→ 22:02, 20 December 2022 (UTC)
- I couldn't find any Polk High Schools either, just North Polk and Southeast Polk. And there don't seem to be any Andrew Johnson High Schools, with or without a cardinal direction, so obviously there are no intersections. (I'm also not holding my breath for a Trump or Nixon High School.) Clarityfiend (talk) 05:59, 21 December 2022 (UTC)
- At least there was a Trump University. —Tamfang (talk) 06:01, 23 December 2022 (UTC)
- There are none in List of educational institutions named after presidents of the United States. Clarityfiend (talk) 06:01, 21 December 2022 (UTC)
- Have you searched on Google Maps, Here.com or OpenStreetMap? They may be better suited for physical locations. --Error (talk) 00:44, 22 December 2022 (UTC)
- I found a Polk County High School in my state (NC), but the county is named after someone else. There's another Polk County HS in Tennessee, but I don't know who that county is named for.--User:Khajidha (talk) (contributions) 00:51, 24 December 2022 (UTC)
- Update: Polk County in Tennessee was named after President Polk before he attained that office.--User:Khajidha (talk) (contributions) 00:53, 24 December 2022 (UTC)