Method and system for voiceprint model training
A model training and voiceprint technology, which is applied in the field of voiceprint model training methods and systems, can solve the problems of reduced robustness of the voiceprint model, and achieve the effects of improving robustness, reducing impact, and excellent performance
Active Publication Date: 2021-10-01
IFLYTEK CO LTD
View PDF4 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004] The technical problem to be solved by the embodiments of the present invention is to provide a voiceprint model training method to solve the problem of robustness of the voiceprint model caused by training a better voiceprint model on the training data with impure user labels in the prior art. Sexually reduced technical issues
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0058] In order to make the above objects, features and advantages of the present invention more comprehensible, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.
[0059] see figure 1 , which is a flow chart of a method for training a voiceprint model provided in an embodiment of the present invention, may specifically include the following steps:
[0060] Step 101: select the voice under multiple user identifications as the training voice;
[0061] In this step, each user ID (ID, Identity) includes multiple voices, and the voices under multiple user IDs are selected as training voices for the voiceprint characterization system. If the voiceprint recognition system for the telephone channel is trained, the voices under 10,000 phone numbers can be selected as the training voice of the voiceprint recognition system.
[0062] Step 102: Determine an initial voiceprint model according to the tr...
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
Login to View More Abstract
The embodiment of the present invention provides a voiceprint model training method and system. The method includes: selecting voices under multiple user identifications as training voices; determining an initial voiceprint model according to the training voices; The training speech under the logo is clustered to obtain the training speech after sorting; the training speech under each user logo after sorting is divided into the first subset, the second subset and the third subset; the training speech under all user logos The first subset is merged into the first training subset, the second subset is merged into the second training subset, and the third subset is merged into the third training subset; subset, the second training subset and the third training subset to obtain the final voiceprint model. The embodiment of the present invention effectively improves the robustness of the voiceprint model, making the performance of the voiceprint system more superior.
Description
technical field [0001] The invention relates to the technical field of voiceprint recognition, in particular to a voiceprint model training method and system. Background technique [0002] The voiceprint feature is one of the important biological characteristics of the human body and has strong individual specificity. It is often used in voiceprint recognition, voiceprint authentication and other fields as a feature of identity authentication. Voiceprint recognition or voiceprint authentication needs to use the pre-trained voiceprint model to extract the feature vector representing the user information from the voice of the confirmation object and the voice to be confirmed, and then use the two feature vectors to calculate the similarity, and finally judge whether for the same user. In the training phase of the voiceprint model, a large number of marked user voices are required as supervised training samples to guide the model to complete the training. However, in many voic...
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
Login to View More Patent Type & Authority Patents(China)
IPC IPC(8): G10L17/04G10L17/20
CPCG10L17/04G10L17/20
Inventor 方昕刘俊华魏思
Owner IFLYTEK CO LTD