The Effect of Noise on Computer-Aided Measures of Voice: A Comparison of CSpeechSP and the Multi-Dimensional Voice Program Software Using the CSL 4300B Module and Multi-Speech for Windows☆
Accepted 13 May 2002.
Abstract
Summary: The effect of noise on computer-derived samples of voice was compared across three different hardware/software configurations. The hardware/software systems included a stand-alone A/D converter (CSL Module 4300B) coupled to a custom Pentium PC used in conjunction with the Multi-Dimensional Voice Program (MDVP) software, and a Creative Labs A/D converter coupled to the same custom PC under software control of MDVP/Multispeech and CSpeechSP. Voice samples were taken from 10 female subjects, then mixed with computer fan noise creating three different signal-to-noise (S/N) levels. Mixed signals were analyzed on the three hardware/software systems. Results revealed that fundamental frequency was most resistant to the degradation effect of noise across systems; jitter and shimmer values, however, were more variable across all configurations. Jitter and shimmer values were significantly higher under certain S/N levels for the MDVP 4300B based system as compared to MDVP for Multi-Speech and CSpeechSP. The findings punctuate the need for sensitivity to recording environments, careful selection of hardware/software equipment arrays, and the establishment of minimal recording conditions (>25dBA S/N) for voice sampling and analysis using computer-assisted methods.
∗University of Wyoming, Division of Communicative Disorders, Laramie, Wyoming, USA
†Speech Science Laboratory, University of Northern Colorado, Greeley, Colorado, USA
Address correspondence and reprint requests to Cecyle Perry Carson, Associate Professor, University of Wyoming, Division of Communicative Disorders, P.O. Box 3311, Laramie, WY 82071-3311, USA
☆ Presented in part at the annual convention of the American Speech-Language-Hearing Association, Washington, DC, 2000.
ABSTRACT