Just another way to show what FT can't: Take a year as a period and now plot the signal of a photo resistor. You will find a frequency of 365 and harmonics! So the day's frequency is 1/365 and a little 1/730 plus a little of 1/1095 and so on. Maybe these harmonics are the reason, why we sometimes feel good?
Just another way to show what FT can't: Take a year as a period and now plot the signal of a photo resistor. You will find a frequency of 365 and harmonics! So the day's frequency is 1/365 and a little 1/730 plus a little of 1/1095 and so on.
Of course there will be some algorithm to ignore un-important harmonics, so everything is beautiful.
Did you get the point? The light signal from a photo sensor will not be periodical to 365. So every 4 years at midnight you will see the stars in a position a little different and after 365*4 periods the stars will be in the same place again. When you sample the daylight over 24hs you can see a basic frequency of one and then harmonics. That is, you describe the light as an average light (the DC component) and then there is another light source that can contribute light (at day) or remove light (at night) and as the sunlight doesn't shine sinuidal, you have to add other light sources with 12, 8, ... hours period.
What I show here: obviously there are usecases, where a fourier transform can describe a behavior without having any realistic background. Voice recognition is such an application, but not as obvious. For this reason, much effort is invested into the wrong place.
Like insisting in Interrupts for the parallel Propeller. They are not needed in that paradigm. But can be used if available. To peace those, that insist in interrupts and to please and ease those, that don't.
And now imagine voice recognition that can follow what I just wrote!
again: yes and no. Yes, I analyse sounds, e.g. heart beats, no, I'm not an expert, as a see experts suffering ;-) basic knowledge. I found, whenever I foutran a signal with spikes, the spikes vanish and new characteric signals appear. That means: if you are looking for an sinusoid of 10 hz in a 1 second intervall and there is a spike, you remove the frequencies > 15Hz successfully and the spike will appear as a 10hz signal. ;-) Whenever you expect something to happen, you have to look for this and to prevent wishfull thinking. This is true for signal voice recognition to. I currently wait for P2 and will use it for such purpose
Comments
https://www.mouser.com/ProductDetail/Microchip-Technology/ATSAMA5D21C-CU?qs=sGAEpiMZZMvu0Nwh4cA1wUll2m7DCiig4IvY9CF2UAAN081UPCTwmQ==
I hope the P2 would make it easier to develop such a concept, even if it wasn't used in production. Kind of like how an FPGA gets used.
P2 is the better candidate here because it is easier to solder than the BGA.
Of course there will be some algorithm to ignore un-important harmonics, so everything is beautiful.
What I show here: obviously there are usecases, where a fourier transform can describe a behavior without having any realistic background. Voice recognition is such an application, but not as obvious. For this reason, much effort is invested into the wrong place.
Like insisting in Interrupts for the parallel Propeller. They are not needed in that paradigm. But can be used if available. To peace those, that insist in interrupts and to please and ease those, that don't.
And now imagine voice recognition that can follow what I just wrote!