The effect of MPEG audio compression on multidimensional set of voice parameters

Abstract

The MPEG-1 Layer 3 compression schema of audio signal, or commonly known as mp3, has caused a great impact in recent years as it has reached high compression rates while also conserving a high sound quality. Previous listening tests have shown that music and speech samples compressed at high bitrates are virtually indistinguishable from the original samples, but very little is known about how compression acoustically affects the voice signal. In Experiment 1 the spectral composition of original and compressed speech signals were analyzed by means of the Long-Term Average Spectrum using the Computerized Speech Laboratory (Kay Elemetrics Corp. (Pine Brook, NJ, USA)). In Experiment 2 a set of 29 voice parameters extracted by using the Multidimensional Voice Program of Kay are compared between original and compressed voices at different bitrates. Results show a high fidelity at high-bitrate compressions (128 and 160 kbit per second (kbps)) both in voice parameters and the amplitude-frequency spectrum. Compressions at 64 kbps or lower bitrates introduces substantial modifications in the voice signal, seriously altering parameters related with tremor, amplitude perturbation, noise, subharmonics and voice irregularities, likewise the signal spectrum is altered in its high frequency region.

• Voice Parameters Mdvp Mpeg Mp3 Long-TERM Average Spectrum Voice Analysis Speech Compression