Dual-sensor voice enhancement method based on statistics model and device

A statistical model and speech enhancement technology, applied in speech analysis, instruments, etc., can solve the problems of statistical characteristic interference, enhanced speech signal-to-noise ratio reduction, speech enhancement effect is not obvious, etc.

Active Publication Date: 2016-06-01
SHENZHEN VOXTECH CO LTD
View PDF12 Cites 14 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

But the speech enhancement method of existing air-conduction sensor and non-air-conduction sensor combination also has the following deficiencies: (1) air-conduction sensor speech and non-air-conduction sensor speech are usually recovered independently, and then the recovered Speech fusion fails to make full use of the complementarity between the air conduction sensor voi

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
  • Dual-sensor voice enhancement method based on statistics model and device
  • Dual-sensor voice enhancement method based on statistics model and device
  • Dual-sensor voice enhancement method based on statistics model and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0090] This embodiment discloses a dual-sensor speech enhancement method based on a statistical model, and the specific process steps refer to the appended figure 1 As shown, it can be known that the dual-sensor speech enhancement method includes the following process steps:

[0091] Step S1: Synchronously collect clean air conduction training speech and non-air conduction training speech, establish a joint statistical model for classification, and calculate the linear spectrum statistical model of air conduction speech corresponding to each classification, which can be further divided into the following step, the process is as figure 2 Shown:

[0092] Step S1.1: Synchronously collect clean air conduction training speech and non-air conduction training speech and divide them into frames, and extract the characteristic parameters of each frame of speech;

[0093] In the above embodiments, the voice receiving module is used to collect clean and synchronized air conduction tra...

Embodiment 2

[0175] The second embodiment discloses a model-based dual-sensor speech enhancement device, which consists of a speech receiving module, a speech statistical model training module, an air conduction noise statistical model estimation module, an air conduction detection speech filter enhancement module, a speech mapping module, and a speech fusion Enhanced modules are composed together, and its structure is as follows figure 2 shown.

[0176] Among them, the voice receiving module is used to synchronously collect clean air conduction training voice and non-air conduction training voice;

[0177] Wherein, the speech statistical model training module is used to establish the joint statistical model and the air conduction speech linear spectrum statistical model;

[0178] Among them, the air conduction noise statistical model estimation module detects the endpoint of the air conduction detection voice, and then uses the pure noise segment of the air conduction detection voice to...

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 discloses a dual-sensor voice enhancement method based on a statistics model and a device. The method comprises steps of combining non-air-conduction sensor voice with air-conduction voice to construct a voice combined statistic model for classification and performing terminal detection, calculating the optimal air-conduction voice filter through the classification result of the combined statistic model, performing filtering enhancement on the air conduction voice, converting the non-conduction voice to the air conduction voice with the air-conduction characteristic through the mapping model, performing weight fusion in order to improve the voice quality. The method adopts the two-stage voice enhancement structure method. When the filtering effect of the air conduction voice is not good because of the strong noise, the second stage voice enhancement performs adaptive weight fusion on the mapping voice filtering voice of the filtering voice and the non-air-conduction voice, so that the better voice enhancement effect can be obtained under the strong noise environment. The method disclosed by the invention can be widely applicable to the video communication, the vehicle-mounted telephone, the multimedia classroom, the military communication, etc.

Description

technical field [0001] The invention relates to the field of digital signal processing, in particular to a statistical model-based dual-sensor speech enhancement method and device. Background technique [0002] Communication is an important means of communication between modern people, and voice is the most common form of communication system, and its quality directly affects the accuracy of people's access to information. In the process of speech transmission, it is inevitable to be interfered by various environmental noises, and its sound quality and intelligibility will be significantly reduced. Therefore, in practical applications, speech enhancement technology is often used to process speech in noisy environments. [0003] Speech enhancement technology can extract useful speech signals from the noise background, and is the basic means to suppress and reduce noise interference. The traditional speech enhancement object is based on speech signals collected by air conduct...

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): G10L21/0272G10L21/038
CPCG10L21/038G10L21/0272Y02T90/00
Inventor 张军陈鑫源潘伟锵宁更新冯义志余华季飞陈芳炯
Owner SHENZHEN VOXTECH CO LTD
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