This repository contains a MATLAB demo that demonstrates the basic principles of perceptual audio coding using MDCT transform and Huffman coding. The MDCT gets applied to the blocked audio signal via polyphase decomposition matrices. The resulting coefficients get quantized by computing the auditory masking threshold and scaling such that the quantization error stays below. The quantized coefficients get then stored using Huffman coding, resulting in a binary file that can be decoded using the Huffman table and inverse scaling coefficients. The resulting binary file is about 15% of the size of the original wave file for the provided examples. Developed during the course work for the Audio Coding lecture at TU Ilmenau, 2012.
The main script audio_coder_demo.m calls all successive steps on an example file and produces the decoded output wave. Original and decoded wave files can be found in './data'.
An example masking threshold for one spectral frame looks like this:
Example of original and decoded waveforms:
Audio Coding, lecture course at TU Ilmenau.