A method and device for improving wake-up performance

A device and performance technology, applied in the field of improving wake-up performance, can solve problems such as poor wake-up performance, and achieve the effect of improving model discrimination ability and improving wake-up performance

Active Publication Date: 2022-06-24
BEIJING UNISOUND INFORMATION TECH +1
View PDF9 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] The invention provides a method and device for improving wake-up performance, which can solve the technical problem of poor wake-up performance in the prior art

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
  • A method and device for improving wake-up performance
  • A method and device for improving wake-up performance
  • A method and device for improving wake-up performance

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0040] An embodiment of the present invention provides a method for improving wake-up performance, such as figure 1 or as shown in 2, including the following steps:

[0041] Step 101: Obtain a general benchmark model, wherein the general benchmark model is obtained by training a neural network with annotated voice; specifically, as image 3 shown, a neural network is trained using a large number of annotated speech as a general benchmark model.

[0042] Step 102: Acquire wake-up voice data of users of different age groups; specifically, the age groups include: children's age group, young and middle-aged age group, and elderly age group; the wake-up voice data covers all ages of the user part.

[0043] Specifically, by initiating a wake-up voice collection task, the target customer groups are divided into different age groups according to age, namely children, young and middle-aged people, and the elderly. Voice collection needs to cover different age groups, for example, ch...

Embodiment 2

[0055] Embodiment 2 of the present invention also discloses a device for improving wake-up performance, such as Image 6 shown, including:

[0056] The first obtaining module 201 is used to obtain a general benchmark model, wherein the general benchmark model is obtained by training a neural network with annotated voice;

[0057] The second acquisition module 202 is used to acquire wake-up voice data of users of different age groups;

[0058] A customization module 203, configured to generate a customized model according to the user of each age group through the general reference model and the wake-up voice data of the user;

[0059] The combining module 204 is configured to combine the customized models of the users of each age group to generate a customized model of all age groups, so as to perform a wake-up operation through the customized models of all age groups.

[0060] In a specific embodiment, the age group includes: the age group of children, the age group of young...

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 present invention relates to a method and device for improving wake-up performance. The method includes: obtaining a general reference model, wherein the general reference model is obtained by training a neural network with marked speech; obtaining wake-up speech of users of different age groups data; for the users of each age group, a custom model is generated through the general benchmark model and the wake-up voice data of the user; the custom model of the user of each age group is combined to generate a custom model for all age groups, to Wake up with the all-ages custom model. By customizing different neural network models for people of different age groups, the ability to distinguish models of different age groups can be improved in a targeted manner, and the wake-up performance of people of different age groups can be improved; Reusing most of the neural network calculations has a limited increase in computing resource consumption.

Description

technical field [0001] The present invention relates to the technical field of speech recognition, in particular to a method and device for improving wake-up performance. Background technique [0002] In the voice interaction system, wake-up is an important interaction entry. Wake-up requires real-time voice processing, not only high wake-up rate and low false wake-up rate, but also demanding system resources such as memory and power consumption. The current wake-up system usually uses a neural network as the acoustic model, and inputs the output of the acoustic model into the wake-up system to obtain the wake-up output and the confidence of the wake-up output. Finally, according to the preset threshold, the wake-up result is obtained. [0003] However, due to resource constraints, the neural network parameters used in wake-up are usually small, and the use of a unified neural network model cannot distinguish the acoustic output of different age groups, resulting in poor f...

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 Patents(China)
IPC IPC(8): G10L15/06G10L15/16G10L15/22
CPCG10L15/063G10L15/16G10L15/22G10L2015/223
Inventor 许东星曹昊周雷梁家恩
Owner BEIJING UNISOUND INFORMATION TECH
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