First things first, thought I'd just mention the latest update to my blog, on the right hand side you can now see a list of what needs to be done in order to complete my prototype. Highlighted in green are the elements that are complete, highlighted in orange are the elements I'm currently working on. So you can keep track of my progress. And this leads in to my current problem.
In processing there are two options for GUI's ( Graphical User Interfaces ).
1) Use a GUI Library
2) Code it all yourself from scratch
Option 2 would take a *long* time, so this leaves option 1. However, even option 1 has problems. Processing has 4 GUI Libraries ( that I am aware of ).
- controlP5
- MyGUI
- Interfascia
- SpringGUI
They all work fine by themselves. However add P3D or OpenGL, and they tend to break. MyGUI and Interfascia seem to die all together. SpringGUI works with P3D but not OpenGL, and controlP5 works with openGL, however the 3D elements always draw on TOP of the GUI. I tried running the GUI elements in a separate window, however that caused issues, and it seems to only work part of the time.
I have heard rumors of a way of re-coding MyGUI to work in conjunction with OpenGL, but I am not 100% sure of its authenticity ( or on the re-compiling process, I'd have to brush up on that ).
So I've gone to the processing forums for support, I'm hoping someone there can offer a suitable solution - Until then, I'm sort of forced to put everything on hold - so hopefully I'll get a solution before then.
If I can't get a solution, I will have to work out a possible solution. Even if this means ensuring no 3D elements move closer than 0 on the Z axis, and drawing all the GUI elements on that point. This will mess up perspective, but it may be the only compromise open to me.
- Anthony
Showing posts with label processing. Show all posts
Showing posts with label processing. Show all posts
Monday, September 3, 2007
Saturday, September 1, 2007
Audio Analysis and Beat Detection version 2
I was bored yesterday afternoon, so I sat down and re-coded from scratch my beat detector. I received an email from Robert Hodgin ( aka Flight404 ) on some tips to getting beat detection working effectively.
His short answer was it's very tricky. He's still trying to figure out the best way to do it himself, and throughout the googleverse it's about the same story wherever you go. The best way to do it I've decided is a version very similar to his own :
Basically you divide the FFT stream up into X number of segments ( I've chosen 6 ), the sizes get larger as they move to the right of the stream ( bass to treble ). For each of these segments ( or "Zones" ), you then calculate an average value, and a "threshold" value. If the average moves higher than the threshold, then you have a dramatic increase in that zone, and possibly ( I say possibly ) a beat.
Talking is boring, how about I show you.
Windows Users : Click here ( .rar 1.4mb )
Mac Users : Click here ( .rar 1.3mb )
Unrar the files. The instructions are in the mac .rar file ( there's a readme.txt ). The program listens to your microphone. Windows users, you can go to your recording preferences and set your recording input to equal your audio output and get direct sound from winamp or wmp or whatever you use. Mac users ? no idea ( if someone figures out how to do it, tell me please !).
Let me explain what you're seeing :
The Red lines are the zone "averages". The yellow lines are the "threshold" values. When 2 or more areas trigger at the same time, I'm saying that's a "beat", the screen will flash white ( very quickly ). You'll notice it's still very jumpy, and requires a LOT of tweaking ,but I think I'm getting there. It's kinda cool though.
Watching it, I've noticed certain trends. In the next version, I'm going to experiement with the timing of the thresholds. How much higher than the current value it has to be to trigger a threshold change, and maybe even "weighting" certain zones to be worth more than others ( as they seem to contain "beat" related frequencies more often than others ... ) Suggestions ?
That's it for now. Enjoy.
- Anthony
His short answer was it's very tricky. He's still trying to figure out the best way to do it himself, and throughout the googleverse it's about the same story wherever you go. The best way to do it I've decided is a version very similar to his own :
Basically you divide the FFT stream up into X number of segments ( I've chosen 6 ), the sizes get larger as they move to the right of the stream ( bass to treble ). For each of these segments ( or "Zones" ), you then calculate an average value, and a "threshold" value. If the average moves higher than the threshold, then you have a dramatic increase in that zone, and possibly ( I say possibly ) a beat.
Talking is boring, how about I show you.
Windows Users : Click here ( .rar 1.4mb )
Mac Users : Click here ( .rar 1.3mb )
Unrar the files. The instructions are in the mac .rar file ( there's a readme.txt ). The program listens to your microphone. Windows users, you can go to your recording preferences and set your recording input to equal your audio output and get direct sound from winamp or wmp or whatever you use. Mac users ? no idea ( if someone figures out how to do it, tell me please !).
Let me explain what you're seeing :
The Red lines are the zone "averages". The yellow lines are the "threshold" values. When 2 or more areas trigger at the same time, I'm saying that's a "beat", the screen will flash white ( very quickly ). You'll notice it's still very jumpy, and requires a LOT of tweaking ,but I think I'm getting there. It's kinda cool though.
Watching it, I've noticed certain trends. In the next version, I'm going to experiement with the timing of the thresholds. How much higher than the current value it has to be to trigger a threshold change, and maybe even "weighting" certain zones to be worth more than others ( as they seem to contain "beat" related frequencies more often than others ... ) Suggestions ?
That's it for now. Enjoy.
- Anthony
Friday, August 17, 2007
FFT Research, Project Choice and Processing Issues
Project Choice
I've decided to go with the audio-responsive visuals. I figure it's the most challenging option available for me and so it's therefore the best way to show off my skills for the end of year expo and beyond.
In Saying this, I wanted to get a decent amount of research done into FFT and Frequency analysis.
Processing Problems
Before I start anything, I've got problems. I wanted to start uploading my processing "sketches". However I've encountered a problem. I spent a large portion of time yesterday afternoon attempting to get the sketches to work as a java applet from within a browser - no joy there, and now I've found that something has broken, and I can't even export as a standalone application - BIG PROBLEM.
The error I get is : "File sonia_v2_9.jar,JSynClasses.jar,JSynV142.dll,libJSynV142.jnilib does not exist"
If anyone knows anything about processing and can help me, I'd appreciate it. I've also posted for support on the processing website here. Until I get this fixed, I can't show you where I'm up to in my experiments - as a last resort I'm sure I can export them from my laptop, so I will have to try that later.
FFT Research
Now to the intersting part ...
I've been doing some research into FFT and Spectrum Analysis. Just so I know what I'm dealing with here. I suppose I should start from the beginning.
In order to work with sound, I need to understand the format in which I can access it. Using processing, and one of the various sound libraries ( In my case, I am using 'Sonia' I can access the live stream of music via my soundcard. This data is intepreted via FFT ( More on this later ), and then outputted into a spectrum array. This array can then be accessed and the music or sound can be intepreted.
So firstly, What is FFT ?
FFT ( or Fast Fourier Transform ) is an "Algorithm to compute the "Discrete Fourier Transform" and it's inverse." Ok, so what is a Fourier Transform ?
The Fourier Transform Produces a Spectrum. It saves the Magnitidue of each frequency component. This data can then be represented as a graph or plot.
So how does this tie in with processing ?
The FFT Data is collated and stored in the array. Each array element is an "FFT Bin". These "bins" are basically storage elements for a *collection* of frequency ranges - not a single frequency. So how are you meant to know what frequencies are contained in which bins ?
Say you have a file with a Sample Rate of 44100Hz. A frequency range of 22050Hz is equally split into 512 bins. Each bin is : 22050 / 512 : 43.066Hz Wide.
The Left most bin ( Lowest Frequency ) has *half* the width of all other bins. ( contains frequencies 0.000Hz to 21.5Hz ).
So what does all this mean ?
This means that I can get data from a sound source, live streaming sound data - examine the frequencies, and calculate which elements represent which aspects of music ( kicks, snares, synths etc ).
And where will this lead ?
Eventually with a sufficiant understanding of the backbone behind the software, and an understanding of the data input and how it is aquired, I can isolate parts of my motion system to react to different parts / types of music in different ways. A greater understanding will help me to create a project of greater impressivenessnessness.
Any Questions ? Any Comments ? Any Suggestions ?
Where now ?
So what I have to do now, is try and get processing working. From there, the experiments continue.
- Anthony
I've decided to go with the audio-responsive visuals. I figure it's the most challenging option available for me and so it's therefore the best way to show off my skills for the end of year expo and beyond.
In Saying this, I wanted to get a decent amount of research done into FFT and Frequency analysis.
Processing Problems
Before I start anything, I've got problems. I wanted to start uploading my processing "sketches". However I've encountered a problem. I spent a large portion of time yesterday afternoon attempting to get the sketches to work as a java applet from within a browser - no joy there, and now I've found that something has broken, and I can't even export as a standalone application - BIG PROBLEM.
The error I get is : "File sonia_v2_9.jar,JSynClasses.jar,JSynV142.dll,libJSynV142.jnilib does not exist"
If anyone knows anything about processing and can help me, I'd appreciate it. I've also posted for support on the processing website here. Until I get this fixed, I can't show you where I'm up to in my experiments - as a last resort I'm sure I can export them from my laptop, so I will have to try that later.
FFT Research
Now to the intersting part ...
I've been doing some research into FFT and Spectrum Analysis. Just so I know what I'm dealing with here. I suppose I should start from the beginning.
In order to work with sound, I need to understand the format in which I can access it. Using processing, and one of the various sound libraries ( In my case, I am using 'Sonia' I can access the live stream of music via my soundcard. This data is intepreted via FFT ( More on this later ), and then outputted into a spectrum array. This array can then be accessed and the music or sound can be intepreted.
So firstly, What is FFT ?
FFT ( or Fast Fourier Transform ) is an "Algorithm to compute the "Discrete Fourier Transform" and it's inverse." Ok, so what is a Fourier Transform ?
The Fourier Transform Produces a Spectrum. It saves the Magnitidue of each frequency component. This data can then be represented as a graph or plot.
So how does this tie in with processing ?
The FFT Data is collated and stored in the array. Each array element is an "FFT Bin". These "bins" are basically storage elements for a *collection* of frequency ranges - not a single frequency. So how are you meant to know what frequencies are contained in which bins ?
Say you have a file with a Sample Rate of 44100Hz. A frequency range of 22050Hz is equally split into 512 bins. Each bin is : 22050 / 512 : 43.066Hz Wide.
The Left most bin ( Lowest Frequency ) has *half* the width of all other bins. ( contains frequencies 0.000Hz to 21.5Hz ).
So what does all this mean ?
This means that I can get data from a sound source, live streaming sound data - examine the frequencies, and calculate which elements represent which aspects of music ( kicks, snares, synths etc ).
And where will this lead ?
Eventually with a sufficiant understanding of the backbone behind the software, and an understanding of the data input and how it is aquired, I can isolate parts of my motion system to react to different parts / types of music in different ways. A greater understanding will help me to create a project of greater impressivenessnessness.
Any Questions ? Any Comments ? Any Suggestions ?
Where now ?
So what I have to do now, is try and get processing working. From there, the experiments continue.
- Anthony
Monday, August 13, 2007
Project Extrapolation
So I thought it was about time I actually posted here explaining what it is I intend to do for my project this semester. As it still yet stands, I am not 100% sure on what I want my project to become, all I know is what I want to work with, and how I want to work with it.
1 ) I want to work with sound and video
2 ) I want to work in processing
3 ) I want to work in 2d and 3d
Using those three parametres, I think currently what I'd like to make are sound-reactive dynamic 3d visuals. Or a least one set of visuals.
Ok, so that's a lot of work. The first question many people would ask is am I capable of achieving this ? Yes. I think I am. I've spent the holidays and last few weeks brushing up on my programming skills ( I've still got a bit of a way to go, but that's what the next few days / weeks are for ), and I've started doing sound experiements in processing.
I'm going to bring a copy of my very simple version 1 beat detector into class tomorrow to see what the tutors and lecturers think of my idea and possible venues for project ideas. Tomorrow I will look into getting web hosting for the .java files and videos. I may end up using vimeo and my optusnet hosting space, but we will see what becomes available to me.
I'm looking forward to this semester, it will be good to have full control and not have to rely on unreliable group members. Let's hope things move smoothly from here.
Subscribe to:
Posts (Atom)
