Current methods for assessing voice quality in patients with voice disorders that rely solely on either auditory perception, or acoustic analysis, are inherently inadequate. There is a pressing need to develop a reliable and objective measure of voice quality can be derived automatically from the acoustic signal, and at the same time is consistent with human perception. The primary aim of this project is to develop an improved (automated) method for objectively measuring the perceptually salient noise in the human voice that can be implemented on existing clinical instrumentation. The new method, named Automated Psychoacoustics-Based Voice-Quality Assessment (APVA), will be based on an innovative combination of well-established principals of human psychoacoustics, and recent advances in signal compression technology. In Phase I of this work the algorithms for estimating a perceptually weighted signal-to-noise (SNR) voice quality measure for continuous speech were developed and validated. Additional pilot testing of the new analysis approach on disordered voices produced highly significant correlations with auditory perception of disordered voice quality. Phase II will be devoted to developing the specific APVA methods and measures that will optimize its performance in the clinical assessment of disordered voices, followed by intergration of the new method into our existing computer speech/voice evaluation systems. At the end of Phase II the new system will be beta tested at independent clinical voice centers. It is expected that the APVA method will lend new insights into the clinical manifestations of disordered voice production, and greatly improve the ability of clinicians to quantify and track the vocal status of voice patients for diagnostic and treatment purposes.
