Research on the mechanism for phonating stressed English syllables based on DIVA model

Abstract

Speech generation and acquisition is a complex cognitive process involving many brain regions. Simulating this process requires establishment of a proper neural network model according to the interaction between sensory organs and motor areas in the brain that occurs during phonation. Directions Into Velocities of Articulators (DIVA) is a neural network model of this type. In the process of speech generation and acquisition, this model can be used to control and simulate the trajectory of articulator and synthesize the desired speech. However, DIVA is flawed in terms of simulating natural language, because it cannot simulate the phonation of English stresses. A DIVA-based neural network computing model for analyzing stresses on English words is proposed. The TD-POSAL (time-domain pitch synchronous overlap add) algorithm is used to tune DIVA׳s speech and rhythm parameters in order to simulate the generation and acquisition of stresses on English words. Simulation showed that the improved DIVA model can represent the features of stressed syllables. The model׳s pitch and duration compensation effect is tested in the increased pitch perturbation experiment. The experiment demonstrates that the improved DIVA model can correctly simulate the generation and acquisition of stresses on English words.