METHOD, DEVICE AND SYSTEM FOR USE OF STATISTICAL INFORMATION TO REDUCE THE NEEDS OF CALCULATION AND MEMORY OF A NEURAL NETWORK BASED VOICE SYNTHESIS SYSTEM.

摘要

A method (400), a device and a system (300) provides, in response to linguistic information, the effective generation of a parametric representation of speech using a neutral network. The method provides, in response to linguistic information, the effective generation of a refined parametric representation of speech, comprising the steps of: A) use a data selection module to retrieve vectors of representative parameters for each description segment depending on the type of phonetic segment and types of phonetic segment included in the adjacent segment descriptions, B) interpolate between the vectors of representative parameters based on segment descriptions and duration to provide statistical parameters interpolated, C) convert interpolated statistical parameters and language information in neural network input parameters, D) using a statistically improved neural network / network neutral with a post-processor to provide the output parameters of the neural network corresponding to a parametric representation of speech, ...