Novel model domain compensation method in remote voice recognition

A technology of speech recognition and compensation method, which is applied in speech recognition, speech analysis, instruments, etc., and can solve problems such as inaccurate compensation parameters and inability to effectively improve the recognition rate

Active Publication Date: 2013-08-21
CHONGQING UNIV OF POSTS & TELECOMM
View PDF3 Cites 17 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In the actual application of speech recognition, the position of the sound source is unknown. In addition, if the position of the sound source changes, but the reverberation compensation in the model domain does not change accordingly, the compensation parameters will become inaccurate, thus The recognition rate cannot be effectively improved

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Novel model domain compensation method in remote voice recognition
  • Novel model domain compensation method in remote voice recognition

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0014] figure 1 A schematic diagram of the principle of the long-distance speech recognition model compensation method is shown. Include steps:

[0015] 1) To calculate the shock response sequence of multiple groups of specific rooms in different locations, the following methods can be used specifically:

[0016] The mirror algorithm is applied to generate multiple sets of random room shock response sequences at different locations. Input the space size parameters of the room, sound absorption coefficient, microphone coordinates and random sound source coordinates. Such multiple parameters are used as a set of parameters. The above parameters are used as input parameters of the mirroring algorithm to calculate the room shock response sequence, and a different sound source coordinates A set of different room shock response sequences will be correspondingly generated.

[0017] Optimize the room shock response sequence, extract energy parameters by frame, and facilitate subseq...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention relates to the technical field of voice recognition and discloses a method and system for eliminating reverberation in remote voice recognition. A novel model domain compensation method in the remote voice recognition comprises the steps: an indoor acoustic environment is simulated and impact response sequences of rooms in different positions are generated through inputting indoor space sizes; clustering analysis is conducted on the generated room impact response sequences, so that the indoor acoustic environment is divided into a plurality of zones, and a corresponding compensation matrix of one impact response sequence of each zone is obtained; in an established recognition network, compensation according to frames is conducted on the recognition network through the compensation matrix of each zone and the optimized compensation is obtained from a plurality of recognition results through the maximum posterior probability thought. Due to the fact that the clustering analysis is conducted on the acoustic environment, model compensation with distinctiveness is conducted on the recognition network, and reverberation resistant performance of the remote voice recognition in the indoor environment is greatly improved.

Description

technical field [0001] The invention relates to the field of speech recognition, in particular to a speech recognition model domain compensation method. Background technique [0002] Speech recognition refers to allowing machines to understand what people say, that is, in various situations, the machine converts human voice signals into corresponding text or commands through recognition and understanding. Its fundamental goal is to develop a machine with hearing function, which can directly accept human speech, understand human intentions, and respond accordingly. From a technical point of view, it belongs to the category of multi-dimensional pattern recognition and intelligent technology. As an interdisciplinary subject, speech recognition is closely related to acoustics, linguistics, artificial intelligence, digital signal processing, pattern recognition and other disciplines, and is widely used in many fields such as industry, military, transportation, and medicine. Wit...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G10L15/20
Inventor 杨勇李劲松
Owner CHONGQING UNIV OF POSTS & TELECOMM
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap