Unlock instant, AI-driven research and patent intelligence for your innovation.

A false wake-up prevention method based on multiple acoustic models

An acoustic model, error prevention technology, applied in speech analysis, speech recognition, unstructured text data retrieval, etc.

Active Publication Date: 2021-02-05
成都启英泰伦科技有限公司
View PDF8 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The solution of the prior art is to expand the garbage lexicon required for training the English acoustic model. This method can solve the false awakening of the English model in the English environment, and can also bring a negative impact on the misidentification of the English model in the Chinese environment. to improve, but still cannot effectively solve the problem

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 false wake-up prevention method based on multiple acoustic models
  • A false wake-up prevention method based on multiple acoustic models

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0027] Specific embodiments of the present invention will be further described in detail below.

[0028] The false wake-up prevention method based on multiple acoustic models of the present invention comprises the following steps:

[0029] S1. Select the language corpus required for training the acoustic models of multiple different languages; wherein the language corpus corresponding to the training of different acoustic models should contain command word corpus;

[0030] S2. Processing the corpus, wherein the pronunciation of the command word corresponding to the command word corpus is the same or similar under multiple language models;

[0031] S3. Using the processed corpus to train the corresponding acoustic models respectively;

[0032] S4. Pack the trained acoustic model and language model to form speech recognition firmware in different languages ​​and burn them into the speech recognition module;

[0033] S5. In the specific recognition process, the voice recognitio...

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

A method for preventing false awakening based on multiple acoustic models and a speech recognition module, wherein the method for preventing false awakening based on multiple acoustic models comprises the following steps: S1. respectively selecting corpora required for training multiple acoustic models of different languages; S2. Process the corpus; S3. Use the corpus to train the acoustic model separately; S4. Package the trained acoustic model and the language model to form speech recognition firmware in different languages ​​and burn it into the speech recognition module; S5. Speech recognition The module inputs the audio to be recognized into multiple voice recognition firmwares simultaneously, and when multiple voice recognition firmwares simultaneously recognize command words, the voice recognition module determines that the audio is a command word and executes the command. The present invention can effectively avoid the misidentification of non-command words as command words due to homophony by training the speech recognition firmware of two different languages ​​to perform speech recognition at the same time, and will not affect the normal recognition of command words.

Description

technical field [0001] The invention belongs to the technical field of speech recognition, and in particular relates to a method for preventing false awakening based on multiple acoustic models. Background technique [0002] With the increasing demand for human-computer interaction, more and more applications related to speech recognition are used in daily life. In the gradually mature voice interaction, people began to pursue more comfort and accuracy of voice recognition. Speech recognition is achieved by matching the language model and the acoustic model. [0003] Since most of the current speech recognition is based on the matching recognition of a single acoustic model and a single language model, there will be a large false awakening situation. For example, in the English model, only part of the English corpus is marked as garbage during training. In the Chinese environment, some Chinese words that are similar in pronunciation to English command words will be misreco...

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): G10L15/00G10L15/06G10L15/183G10L15/26G06F16/31G06F40/117
CPCG10L15/005G10L15/063G10L15/183G10L15/26G06F16/31G06F40/117
Inventor 舒畅何云鹏许兵
Owner 成都启英泰伦科技有限公司