Audio Compression: Psychoacoustics of MP3

Author/Creator

Author/Creator ORCID

Date

2019

Type of Work

Department

Mathematics and Computer Science

Program

Citation of Original Publication

Rights

Abstract

Digital audio compression is used to minimize the size (number of bits) of an audio file in order to transmit the file effectively and clearly. Only necessary data points are encoded which will continue to represent the audio signal well. Examples of audio compression can be found in music and cell phone transmission. The main compression standard studied is MP3 digital audio compression, which is an effective tool for reducing the size of audio files. Reducing file size, however, may put sound quality in jeopardy. I created an audio test and questionnaire to see if humans can distinguish a difference in quality between similar compressed MP3 files. MP3 audio encoding implements a Psychoacoustic Model, which recognizes that most humans cannot perceive frequencies outside of the range 20Hz to 20kHz, and these frequencies are therefore not encoded. The audio test analyzes this technique used in MP3 encoding by compressing portions of a song at various sampling rates (samples per second) while keeping bitrate (number of bits read per second, corresponding to file size) constant at 96kbps (kilobits per second). Sampling rate is the number of samples per second that is read from a continuous signal and converted to a discrete signal, hence the more samples, the more data, and the higher the quality. The greater the bitrate and sampling rate, the higher the audio quality, but correspondingly the greater the storage necessary to save. Subjects are asked to listen and compare two different sound files and then identify which sound they deem has a higher quality. The goal of this project is to attempt to identify an optimal sampling rate for compression that reduces the file size while also maintaining high quality. MP3 encoding removes information from the digital audio file, and this can result in some humans detecting a reduction in the quality of the audio.