Title:
SEQUENCE TO SEQUENCE TRANSFORMATIONS FOR SPEECH SYNTHESIS VIA RECURRENT NEURAL NETWORKS
Document Type and Number:
WIPO Patent Application WO/2018/081163
Kind Code:
A8
Abstract:
A system eliminates alignment processing and performs TTS functionality using a new neural architecture. The neural architecture includes an encoder and a decoder. The encoder receives an input and encodes it into vectors. The encoder applies a sequence of transformations to the input and generates a vector representing the entire sentence. The decoder takes the encoding and outputs an audio file, which can include compressed audio frames.
Inventors:
HALL DAVID LEO WRIGHT (US)
KLEIN DANIEL (US)
ROTH DANIEL LAWRENCE (US)
GILLICK LAURENCE STEVEN (US)
MAAS ANDREW LEE (US)
WEGMANN STEVEN ANDREW (US)
KLEIN DANIEL (US)
ROTH DANIEL LAWRENCE (US)
GILLICK LAURENCE STEVEN (US)
MAAS ANDREW LEE (US)
WEGMANN STEVEN ANDREW (US)
Application Number:
PCT/US2017/058138
Publication Date:
May 09, 2019
Filing Date:
October 24, 2017
Export Citation:
Assignee:
SEMANTIC MACHINES INC (US)
International Classes:
G10L25/00
Attorney, Agent or Firm:
CREASMAN, Jason, C. (US)
Download PDF: