Training method of voice wake-up model, wake-up word detection method and related equipment

A voice wake-up and training method technology, applied in voice analysis, voice recognition, instruments, etc., can solve the problems of high hardware performance requirements, poor stability of voice wake-up methods, etc., and achieve the effect of optimized training

Active Publication Date: 2022-01-21
深圳市友杰智新科技有限公司
View PDF5 Cites 8 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The main purpose of this application is to provide a voice wake-up model training method, a wake-up word detection method and related equipment, aiming to solve the disadvantages of poor stability of existing voice wake-up methods or high requirements for hardware performance

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
  • Training method of voice wake-up model, wake-up word detection method and related equipment
  • Training method of voice wake-up model, wake-up word detection method and related equipment
  • Training method of voice wake-up model, wake-up word detection method and related equipment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0029] In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present application, and are not intended to limit the present application.

[0030] refer to figure 1 , an embodiment of the present application provides a method for training a voice wake-up model, including:

[0031] S1: Acquire sample data and label data corresponding to the sample data, the sample data including wake-up word positive sample voice data;

[0032] S2: the sample data and the label data are input into the neural network as training samples to carry out model training, and in the model training process, the label data is used as the supervised learning target of the sample data, using unified multi-label cross entropy The loss fun...

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 provides a training method of a voice wake-up model, a wake-up word detection method and related equipment. During model training, training of the voice wake-up model is effectively optimized by using a unified multi-label cross entropy loss function in combination with a form of voice phonemes. When wake-up word detection is carried out, multi-label classification of phonemes is executed through the voice wake-up model, and judgment of whether wake-up words exist in wake-up voice data is converted into judgment of whether wake-up word phonemes exist so that wake-up words are detected from finer granularity, the false wake-up rate is reduced, and the model stability is improved. Meanwhile, the voice phoneme sequence is obtained according to the wake-up word phoneme combination existing in the wake-up voice data so that the output of the voice wake-up model is notunnecessarily decoded and phoneme sequence recognition isconverted into judgment whether the phonemes in the sequence exist or not, the post-processing form of the output of the voice wake-up model is greatly simplified, and the data operand and the performance requirement on the terminal equipment are effectively reduced.

Description

technical field [0001] The present application relates to the technical field of speech recognition, in particular to a method for training a speech wake-up model, a method for detecting wake-up words, and related equipment. Background technique [0002] Voice wake-up is the first step in human-machine voice interaction, which is used to start the smart module of the device. Common wake-up models can be divided into two categories according to the model structure, one is the judgment model based on the classification model, and the other is the recognition model based on the sequence prediction model. Although the classification model is simple to implement, it has poor stability; while the sequence prediction model needs to cooperate with a complex decoding mechanism to achieve good recognition results, so the computing power required for deployment hardware is relatively high. Contents of the invention [0003] The main purpose of this application is to provide a voice ...

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
IPC IPC(8): G10L15/02G10L15/06G10L15/16G10L15/22
CPCG10L15/02G10L15/063G10L15/16G10L15/22G10L2015/025G10L2015/223
Inventor 徐泓洋王广新杨汉丹
Owner 深圳市友杰智新科技有限公司
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