MP3s are lines of programming that is known as MPEG-1 Audio Layer 3. In this format, the complexity is at its highest and is able to compress more information into smaller sizes and produce the best audio quality. The first step in the MP3 process is translating the data into codes using the Huffman process. Essentially, the data is compressed to smaller bits called frequency domain samples. The music is essentially translated into information that is recoded and decoded into facts. The second step in the process involves decoding of the information and undergoes a process called requantization. Requantization approximates the values into smaller ratios and MP3 encoders are non-linear quantizers that in essence translate the values. This involves the removal of unnecessary noise from the essential data. The last step is reordering the decoded data into frequencies in time into joint stereo output. This means that the smaller bits of information is again altered and then expanded into music from electronic bits of data. The process is simplified here, but coding and decoding are the essential steps in MP3 music playing.

SOURCE: http://blog.bjrn.se/2008/10/lets-build-mp3-decoder.html