A method for automatic speaker labeling

An automatic labeling and speaker technology, applied in speech analysis, speech recognition, instruments, etc., can solve unrealistic problems, avoid errors, ensure accuracy, and improve accuracy.

Active Publication Date: 2020-05-29
北京灵伴即时智能科技有限公司
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Therefore, the assumption that the role of the speaker is fixed once the speaker is registered is also unrealistic in practical applications.

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 for automatic speaker labeling

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0040] Such as figure 1 As shown, an embodiment of a method for automatic speaker labeling provided by the present invention includes the following steps:

[0041] Step 1: Perform speech recognition on the speech to be marked, obtain the text content and the position of the silent segment, and divide the speech into multiple short sentences according to the position of the silent segment.

[0042] In order to obtain more accurate results, a phoneme-based speech recognition system can be used to obtain the phonemes in the speech and the start and end times of silence.

[0043] Use the silent part to split the speech into multiple short sentences, and use the silent point as a possible speaker change point for subsequent speaker labeling.

[0044] Step 2: Identify the text content of each short sentence, use the pre-trained content-related role model to score the role confidence of each short sentence, and use the sentences whose scores meet the requirements to form the voice d...

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 belongs to the field of speech signal processing, specifically relates to an automatic marking method for a speaker, and aims at improving the marking accuracy of the speaker role. The method comprises the steps: carrying out the speech recognition of to-be-marked speech, obtaining text contents and the position of a silent segment, and dividing the speech into a plurality of short sentences according to the position of the silent segment; carrying out the role confidence scoring of each short sentence through a pre-configured role model based on contents and an above recognitionresult, and forming a corresponding role speech data set through the speech of the short sentences with the high scores; taking the speech in a current role set as the training data, and obtaining aspeaker recognition model of the role; scoring other short sentences through the speaker recognition model, marking the speech of sentences meeting the requirements as the corresponding role, adding the role to the role speech data set, and updating the speaker recognition model; carrying out the speaker recognition of the remaining sentences through the updated model, and obtaining a role markingresult.

Description

technical field [0001] The invention belongs to the field of speech signal processing, and in particular relates to a method for automatically marking speakers. Background technique [0002] Speaker annotation is to mark the relevant attributes of each speaker in the multi-person dialogue voice, mainly including the number of speakers in the voice, the start and end time of each speaker's voice, and the role in the dialogue. Speaker annotation technology has broad application prospects in voice document information mining, mobile interaction, identity verification, audio monitoring and other fields; it is especially suitable for scenarios where the speakers participating in the dialogue have obvious role characteristics, such as voice customer service, court defense, and TV interviews Wait for the voice dialogue scene. [0003] Existing methods perform speaker separation and role labeling on the basis of speaker voice modeling and automatic classification. The premise is a...

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): G10L15/06G10L15/07G10L15/26
CPCG10L15/063G10L15/07G10L15/26G10L2015/0631G10L2015/0635G10L2015/0638
Inventor 庞在虎张志平陈博朱风云
Owner 北京灵伴即时智能科技有限公司
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