A new method was developed for the quantitative evaluation of hoarseness with the use of a personal computer.
The prediction error (residual signal) of voice samples was obtained by a linear predictive coding (LPC) method. The differential of the error was processed statistically by applying the Rayleigh distribution. Then, the coefficient of the probability density function was calculated and used as a parameter for the degree of hoarseness.
This method was applied to 42 normal and 55 abnormal voice samples, and it was proved to be useful for the quantitative evaluation of hoarseness. It was also useful for the objective evaluation of voice improvement after phonosurgery.