titre de mon image

FAQ

Frequently asked questions

Translated by Loïc Corbel



Why to use PMPC ?

PMPC grants the user to obtain quality signals classifications with ease. Studies hardly conceivable with other systems can become possible.

What is PMPC forte ?

An important part of the work is automated and the results are dealt by the computer. This way, repetitive tasks are eliminated like errors linked to data manipulation.
The researcher can truely spend this gained time for reflexion and choice of methodology, as the only required effort is the tag and choice of sequences to compare. A large panel of configurations is available when sequences are compared by Complearn.

I'm using Windows or Mac... ?

Poor you ! Sadly, we can't provide a Windows version as MPTK and Complearn are Unix dedicated software.
For Mac OSX, PMPC should be rather easily adaptable to it.

Have there been any other tries before you ?

Yes, I've been in touch with a researcher who tested atomic decomposition using Guimauve a few years ago. She was still missing a way to analyse the results which appeared a while later with the compressing concept introduced in Complearn.

Why would you succeed where others have failed ?

Taking into account so many softwares dates of publication, the evolution of computering and mathematics, we couldn't think about using such a system before year 2000.
The atomic theory of soundwave appeared during the 1940's with Gabor, but was already underlying to quantum mechanics by the end of the 20th century.

Wavelets and Matching Pursuit have been theorized near year 1985, and softwares started being written from this point on. Complearn appeared in 2003. Without the synergy of all those theories and softwares, PMPC would never have been achieved.

What have been your working method ?

For my thesis work I didn't have to comply with a laboratory project not to be greedy for results, which gave me the opportunity to challenge well established principles and interest myself to little known results and theories.

Besides, I often made mistakes on concepts or exaggerated their impact, which allowed me to advance and end up with a working prototype.

The ideas of atoms, then groups of atoms or molecules oriented me to data processing coming from the biology field of research, Complearn being a nice outcome of it. The idea of music note and score of the signal grant me to have great confidence if the process.

Has PMPC been tested on many cases ?

Sadly, not yet. I'm not working in linguistics anymore, and am close for retirement. I'm missing time to exploit all the possibilities of PMPC. I'm hoping that researchers, attaining limits of FFT will pursue my work. It seems like there is no theoretical way for this analysis chain.

Why all those very complicated steps while we could, according to
Complearn documentation, simply compress the soundfiles
themselves ?

Because contrary to what one might think, the numerical file of a signal contains very little data, and only computer data. In fact, it only contains the required informations to reproduce and image of the original signal.
We maintain that it is at maximum entropia level for reproduction of the signal.

Some will say that the digitized signal contains so many datas that we are unable to understand it all ! This is a deception that is easy to demonstrate.

A music score gives, in a highly synthetical form, complete datas on a tune. Notes, their pitch, duration, envelope, tone, informations concerning rythm and even interpretation according to my analyses. MPTK provides high level datas on the signal's music in opposition to the low level data of the digitized file.

On digitized files, compressing methods give random results including « compressed » files larger than the original, which proves that our assertion concerning their internal disorder or maximum entropia is true. Some specialized compressing softwares (FLAC) only work, more or less, on music. They fail on voice compression for example.

The series of values inside a digitized file are not human language. Music is, and atomic decomposition produces a language close to the musical one. (A. Grossmann).

Working using digitized files as a basis should be forbidden, and that will always remain the same.

What are the points that give you confidence in your system ?

1- The atomic representations which gives a more detailed and controlled description of the signal compared to Fourier methods; ;

2- Atoms which are like music notess, the group of atoms which can be considered as a molecule or the musical score of the signal ;

3- Music score comparison, notably using Complearn, which gave superior results to any other method, for musical genre, composer or performer detection ;

4- The fact that only classifications that gave experimentally reliable results were based on written scores or MIDI files (MIDI files are written scores translated in computer language).).

Choice of atoms and the shape of their wave (Guimauve)

exemple