The MP3 Paper

Note: This first draft is more of an outline. Most of my writing so far has been code, thus the demo is mostly up and running but the rest is yet to come.

Introduction

So You Think You Know MP3…

In this paper, I aim to describe the mp3 format by decoding the psychology that it employs. I’ll start with a brief history of audio recording technologies and the practice of audio coding, including the innovations behind perceptual audio coding. The subsequent demonstration section is an interactive exhibition of three critical steps in the mp3 algorithm, where you can upload an uncompressed (.wav) audio file and see the perceptual coding in action.

The latter sections (order TBD) expand the scope of the paper to more philosophical horizons, contemplating the current ideologies surrounding lossy audio compression, as well as the speculative future of coding paradigms (i.e. touching on AI-assisted compression).

If you’re curious, here’s the code.

Part One

A Brief History

As we know, sound waves propagate through air and interfere additively, such that every excitation of air from any source, no matter its strength, contributes to a single perceived stream of sound which is, “in short, a tumbled entanglement of the most different kinds of motion, complicated beyond conception [1, p. 26].”

To be discussed:

Part Two

Demonstration

Helmholtz was correct in declaring the mechanisms of wave composition complicated, though via the progression of his own work, and the nearly two centuries of science that succeeded it, the qualifier ‘beyond conception’ is proven hyperbole. In this section, I hope to distill some of the wisdom attained in those years into an interactive demo of the mp3 algorithm, blown up step-by-step with helpful visualizations. The goal is to show the nature of the data that’s being discarded, and of the data that’s being kept around.

Step One

Upload File



No File Selected

Now you may adjust the bit-rate. Its like a quality slider. Choose your bit rate now!

Step Two

Adjust Bit Rate


128 kbps


Algorithm: Polyphase Filtering

We run the data through a custom filterbank to divide the audio signal into 32 equal-width frequency subbands. For the visualization, We show the subband split visually, alongside other stats from this step in the encoding process.


Algorithm: Psychoacoustics Modeling

Masking is a primarily neurological phenomenon wherein “louder sounds hide, i.e., ‘mask’, softer sounds in their spectral or temporal vicinity” [2]. That is to say, masking causes us to lose track of the fainter sounds when a loud sound is introduced. Certain signals are better maskers than others (white noise, for instance). This part of the mp3 algorithm involves ‘replacing’ weaker signals with noise (random aperiodic signals). This effect is essential for perceptual coding algorithms to work properly.

Psychoacoustics has examined the concept of auditory masking and its effect on compression. The human ear has ‘critical subbands’ where masking occurs more conspicuously. This visualization will show some of that subband analysis in a digestible way. In this step, each visualization should be quite unique to the audio file that we’re encoding.


Algorithm: Bit Allocation

The bit allocation uses information from the psychoacoustic model to determine the number of code bits to be allocated to each subband. This process can be described using the following formula: MNRdB = SNRdB - SMRdB. For this visualization, we want to build off the psychoacoustics results and show which bands are getting more bits.


Algorithm: Result

Listen to the resulting mp3 file:


Now try again with a different bitrate…

Parts Three, Four etc.

Rest of the Paper

paper pending…

Subjects to discuss: