Invention Grant
- Patent Title: Multi-class constrained maximum likelihood linear regression
- Patent Title (中): 多类约束最大似然线性回归
-
Application No.: US12114005Application Date: 2008-05-02
-
Publication No.: US08386254B2Publication Date: 2013-02-26
- Inventor: Neeraj Deshmukh , Puming Zhan
- Applicant: Neeraj Deshmukh , Puming Zhan
- Applicant Address: US MA Burlington
- Assignee: Nuance Communications, Inc.
- Current Assignee: Nuance Communications, Inc.
- Current Assignee Address: US MA Burlington
- Agency: Sunstein Kann Murphy & Timbers LLP
- Main IPC: G10L15/20
- IPC: G10L15/20 ; G10L15/02
Abstract:
A method of speech recognition converts an unknown speech input into a stream of representative features. The feature stream is transformed based on speaker dependent adaptation of multi-class feature models. Then automatic speech recognition is used to compare the transformed feature stream to multi-class speaker independent acoustic models to generate an output representative of the unknown speech input.
Public/Granted literature
- US20090024390A1 Multi-Class Constrained Maximum Likelihood Linear Regression Public/Granted day:2009-01-22
Information query
