Stochastic models of pitch jitter and amplitude shimmer for voice modification

We present a voice modification algorithm for transforming a modal voice to a hoarse voice. The algorithm is based on modifying the jitter and shimmer of the voice, which are long known to be connected to hoarseness. A pitch synchronous linear prediction analysis-resynthesis process is used to increase jitter and shimmer in modal voices, where the data for the jitter and the shimmer is obtained from a stochastic model based on naturally hoarse voices. A formal evaluation of the algorithm conducted with multiple voices and listeners showed that the produced voices are perceived as close to natural hoarseness, suggesting the applicability of the algorithm to various voice modification and voice synthesis tasks.