Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: thesis on lossless audio compression (Read 7228 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

thesis on lossless audio compression

Hello everyone, i'm new and this is my first topic, so if i make errors please notice me.
I'm an information engineering student of Padova University (italy).

I've to do a thesis about lossless compression, which includes two part,
1) general info about how it works and state of art of algorithms

2) .....

the part 2 is free, for simulation, test some code for compression...

I think it will be nice to do simulation to compare mp3 and flac (for example) audio, with their Fourier Transforms
to evidence the difference between those (i've just made some of this simulations) or evidence that using huffmann coding we use less number of bits .

How i can do this? matlab work well for it??

Any idea to help me is welcome, if you have suggestions for second part write me, also if in your opinion will be nice make some simulation about something you could be interessed!!!!

PS. i've a lot of IEEE publications about this arguments, so about the theory i'm allright, i want to do some practice on signal processing!

Thanks!!!

thesis on lossless audio compression

Reply #1 –
Quote from: panez on
I'm an information engineering student of Padova University (italy).

Greetings from Germany to Italy!

Quote from: panez on
I've to do a thesis about lossless compression, which includes two part,
1) general info about how it works and state of art of algorithms
2) .....
the part 2 is free, for simulation, test some code for compression...

I think it will be nice to do simulation to compare mp3 and flac (for example) audio, with their Fourier Transforms


It's like comparing apples to oranges, isn't it? If you want to write about lossless compression there's no need to do any testing of MP3. I also don't see why you would want to compare "their Fourier Transforms".

Quote from: panez on
to evidence the difference between those (i've just made some of this simulations) or evidence that using huffmann coding we use less number of bits.

How i can do this? matlab work well for it??

First, you gotta figure out something that makes a little more sense. And I would think that this is part of your job. It's probably best if you start with part 1 right away so you know what's going on in the lossless audio compression department. Then, you may get some nice ideas on what to simulate or compare...

Cheers,
SG

thesis on lossless audio compression

Reply #2 –
Quote from: SebastianG on
Quote from: panez on
I'm an information engineering student of Padova University (italy).

Greetings from Germany to Italy!

Quote from: panez on
I've to do a thesis about lossless compression, which includes two part,
1) general info about how it works and state of art of algorithms
2) .....
the part 2 is free, for simulation, test some code for compression...

I think it will be nice to do simulation to compare mp3 and flac (for example) audio, with their Fourier Transforms


It's like comparing apples to oranges, isn't it? If you want to write about lossless compression there's no need to do any testing of MP3. I also don't see why you would want to compare "their Fourier Transforms".

Quote from: panez on
to evidence the difference between those (i've just made some of this simulations) or evidence that using huffmann coding we use less number of bits.

How i can do this? matlab work well for it??

First, you gotta figure out something that makes a little more sense. And I would think that this is part of your job. It's probably best if you start with part 1 right away so you know what's going on in the lossless audio compression department. Then, you may get some nice ideas on what to simulate or compare...

Cheers,
SG



hi!!! thanks for the answer!
you're right when you say that is better to start with part one, i've read a lot of IEEE publications about this argument
and in some of those i've found comparison between mp3 and lossless, or plots about huffman coding coding vs # of bit,
here's why of my idea.

thanks again for the answer, if I will found a really nice idea i will post it here!

thesis on lossless audio compression

Reply #3 –
Comparing MP3 to lossless would be the same as comparing MP3 to the original. Your thesis would be about MP3 (lossy coding) if you took this route. I agree your work on part 1 will probably turn up something interesting for you that you'll want to explore in part 2.

thesis on lossless audio compression

Reply #4 –
yes, you're both right..
my goal in compare that kink of file was only this:
imagine to talk about this argument to someone who don't know nothing, if you
show it that the original file, the lossless file have the same spectrum, instead the mp3 no,
he probably understand better what we are talking about.

obviously that ( i've just make this part in matlab) is only an introduction, not the important part!

thesis on lossless audio compression

Reply #5 –
Comparing spectra may not be the best way to show that the files are or are not the same. Try subtracting one from the other and show the spectrum of the difference. Lossless minus wav will be total silence, while lossy minus wav will have some residual.

Don't take this comparison too far, though. The spectrum and magnitude of the lossy minus wav residual is not useful in judging the quality of the lossy encoding.

thesis on lossless audio compression

Reply #6 –
Quote from: pdq on
Comparing spectra may not be the best way to show that the files are or are not the same. Try subtracting one from the other and show the spectrum of the difference. Lossless minus wav will be total silence, while lossy minus wav will have some residual.

Don't take this comparison too far, though. The spectrum and magnitude of the lossy minus wav residual is not useful in judging the quality of the lossy encoding.



infact i will not take much care from that analysis, it's only a "naif"  demostration!

@antropoid

thanks for the link, i found it very intersting and part of my work alredy done is similar!



after reading some papers i think that i'll analyze different implementation and algorithms of codec's predictors as main theme!

thesis on lossless audio compression

Reply #7 –
Quote from: antropoid on
http://www.stereophile.com/features/308mp3cd/index.html

I don't think this is helpful. And I'm almost sure that this article has been already ripped apart (for its dramatization of artefacts and use of misleading pictures*) by the HA community.

(*)
Quote
However, a picket fence of very-low-level vertical lines can be seen. These represent spurious tones that result, I suspect, from mathematical limitations in the codec.

the last picture clearly shows the output of an mp3 decoder that didn't use dithering. So, the author's suspection is wrong. The issue here is not the mp3 format.

thesis on lossless audio compression

Reply #8 –
Quote from: SebastianG on
Quote from: antropoid on
http://www.stereophile.com/features/308mp3cd/index.html

I don't think this is helpful. And I'm almost sure that this article has been already ripped apart (for its dramatization of artefacts and use of misleading pictures*) by the HA community.


Ok, I think a link for the ripping of this article would be helpful since it can pinpoint issues that must be considered by panez in his thesis.

thesis on lossless audio compression

Reply #10 –
Thanks, It is really helpful indeed.

thesis on lossless audio compression

Reply #11 –
Hello from Latrobe University (Australia)
I've also started a Literature Survey on "Audio Compression schemes"
I'm an Undergrad student, not a masters or PhD. So I'm not expected to change the world. However I would like to know a lot more about the lossless area. I've come to the realisation that pretty much EVER lossless encoding format stems from Shorten (FLAC, monkey audio, Meridian Lossless Packing, Dolby Digital TrueHD) but that they basically add and modify the implementation.

What I would like to do for my thesis is further investigate Block sizes and prediction coefficients in how they relate to REAL data in REAL world. I'm good with mathematics, I understand all the concepts, I just dislike the fact that most of the examples from IEEE are all theory with very abstract examples.
I want to get actual musical samples and be able to say blocking should be done here to here and linear coefficients reduced or increased the residuals as such on these graphs, THEN go into the reasons and theory behind it. I just don't fully understand how channel correlation + linear prediction isn't producing better than ~48% compression, I realise that isn't the whole story.

I'm comfortable with Java and C/C++ (not the best) and have been trying to look into the LibFLAC libraries but it seems to have gotten to the stage where its almost impossible to follow and understand what's happening unless you were in from day 1...And as I've never done any audio compression coding before (my experience has been mainly application programming) its just doing my head in a little...Has anyone got any suggestions as to where to start? even if its not FLAC

Has anyone found useful tools in MATLAB for coding or analysis of WAV files

thesis on lossless audio compression

Reply #13 –
Quote from: ggf31416 on
For a start you may want to read:
http://www.monkeysaudio.com/theory.html
http://www.lossless-audio.com/theory.htm


Thanks for the quick reply but I have read and I understand the concepts. I was asking specifically about the FLAC implementation and how the coding actually works from a WAV file to flac, how it makes its decisions etcetera. I'll post my questions in the FLAC sections as I realized how may I have...
 

thesis on lossless audio compression

Reply #14 –
Quote from: RickJames on
Hello from Latrobe University (Australia)
I've also started a Literature Survey on "Audio Compression schemes"
I'm an Undergrad student, not a masters or PhD. So I'm not expected to change the world. However I would like to know a lot more about the lossless area. I've come to the realisation that pretty much EVER lossless encoding format stems from Shorten (FLAC, monkey audio, Meridian Lossless Packing, Dolby Digital TrueHD) but that they basically add and modify the implementation.

What I would like to do for my thesis is further investigate Block sizes and prediction coefficients in how they relate to REAL data in REAL world. I'm good with mathematics, I understand all the concepts, I just dislike the fact that most of the examples from IEEE are all theory with very abstract examples.
I want to get actual musical samples and be able to say blocking should be done here to here and linear coefficients reduced or increased the residuals as such on these graphs, THEN go into the reasons and theory behind it. I just don't fully understand how channel correlation + linear prediction isn't producing better than ~48% compression, I realise that isn't the whole story.

I'm comfortable with Java and C/C++ (not the best) and have been trying to look into the LibFLAC libraries but it seems to have gotten to the stage where its almost impossible to follow and understand what's happening unless you were in from day 1...And as I've never done any audio compression coding before (my experience has been mainly application programming) its just doing my head in a little...Has anyone got any suggestions as to where to start? even if its not FLAC

Has anyone found useful tools in MATLAB for coding or analysis of WAV files



Hi Rick, i've found flac toolbox and i'm using that for some simulation.
if you think it could be helpful here you can download it
http://www.mathworks.com/matlabcentral/fil...ncodingdecoding.