Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Low frequency logarithmic spectrum based robust feature extraction method

An extraction method and robust feature technology, applied in speech analysis, speech recognition, instruments, etc., can solve problems such as system performance degradation and unavailability, and achieve the effects of small calculation, reduced impact, and improved environmental robustness

Active Publication Date: 2018-11-30
HOHAI UNIV
View PDF14 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, due to the influence of speech variability, the characteristics of the MFCC extracted in the actual environment may be quite different from the training speech, which will lead to a decrease in system performance, or even unusable

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
  • Low frequency logarithmic spectrum based robust feature extraction method
  • Low frequency logarithmic spectrum based robust feature extraction method
  • Low frequency logarithmic spectrum based robust feature extraction method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0019] Below in conjunction with specific embodiment, further illustrate the present invention, should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention, after having read the present invention, those skilled in the art will understand various equivalent forms of the present invention All modifications fall within the scope defined by the appended claims of the present application.

[0020] Such as figure 1 As shown, the robust feature extraction method based on low-frequency logarithmic spectrum mainly includes preprocessing, FFT, logarithmic transformation, low-pass filtering, exponential transformation, Mel filtering, DCT and time-domain difference.

[0021] 1. Pretreatment

[0022] In the speech preprocessing stage, the input speech is windowed and divided into frames to obtain the frame signal x. The sampling frequency of the speech signal is 8000Hz, the window function...

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 low frequency logarithmic spectrum based robust feature extraction method. Feature parameters can be extracted by using a logarithmic spectrum profile of a speech signal. Firstly, logarithmic transformation is performed on an amplitude spectrum of the speech signal to obtain a logarithmic spectrum. Then, the logarithmic spectrum is regarded as a time domain signal, and low-pass filtering is performed on the logarithmic spectrum by using a digital filter to obtain a low-frequency logarithmic spectrum. Finally, exponential transformation, Mel filtering, logarithmic transformation and discrete cosine transformation are performed on the low frequency logarithmic spectrum of the speech signal, and time domain difference is performed to obtain feature parameters of thespeech signal. The method can improve the environmental robustness of the feature parameters of the speech signal, reduce the influence of the speaker change on a speech recognition system, and has the advantages of small calculation amount and easy real-time realization.

Description

technical field [0001] The invention belongs to the technical field of speech recognition, and in particular relates to a robust feature extraction method for performing low-frequency filtering on a logarithmic spectrum of a speech signal and reducing the influence of environmental mismatch on a speech recognition system. Background technique [0002] The acoustic model of each speech unit of the speech recognition system is generally trained with the training speech of several people in a quiet environment. If the training speech can cover the pronunciation characteristics of the actual speaker, the speech recognition system can achieve a high recognition rate. However, the pronunciation methods of speakers in different regions are quite different, and there are too many types of pronunciation methods, it is difficult to consider all pronunciation methods in the training process of the acoustic model. Moreover, if too many different training voices are used in the training ...

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): G10L15/02G10L25/03
CPCG10L15/02G10L25/03
Inventor 吕勇
Owner HOHAI UNIV
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com