Perception of pathological voice quality is essential in clinical voice evaluation and validation of objective measures of voice. Patients and their families often decide whether treatment is successful based largely on how the voice sounds. Similarly, clinicians make many decisions about managing voice disorders based upon perceptual judgments. However, these """"""""subjective"""""""" measures of voice quality are not highly regarded as either clinical or research tools, because of problems with reliability, because they are considered to lack sufficient objectivity, and because there is no accepted standard set of perceptual scales used by clinicians. We hypothesize that problems in voice quality measurement do not originate within the listener, but rather derive from the methods used to measure what listeners hear. Studies completed during the first funding period indicate that listeners share few, if any, perceptual features for pathological voice, demonstrating that traditional voice rating scales represent perceptual categories of questionable validity. Consequently, the proposed research departs from traditional rating scale methods by applying a synthesizer for pathological voice quality to study fundamental issues concerning reliability and validity of voice quality measures. The proposed analysis by synthesis (AbS) method provides listeners with the opportunity to construct a synthetic signal which perceptually matches the natural pathologic voice under observation, explicitly linking an acoustic signal to the perceived voice quality. This method models quality as a whole, avoiding the problems of defining valid, meaningful perceptual scales for complex signals that are perceived in a variable fashion. By modeling the signal that generates a perception, AbS-should also provide a measurement tool that is resistant to listener-related variability. A real-time synthesizer for pathological voices will be constructed and tested by evaluating various source models and the perceptual importance of source-filter interactions, and by modeling a broad range of pathological qualities. Other studies will examine the reliability and validity of AbS as a method for evaluating vocal quality. The long term goal of our research continues to be the development of a voice evaluation protocol to maximize the reliability and validity of voice quality measurement. Once this goal is accomplished, standardization may be achievable. Considering the key role of voice quality perception in both clinical and research practices, the need for increased reliability, validity, and eventual standardization in this field cannot be overstated.
Showing the most recent 10 out of 35 publications