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

A Multi-environment Model Isolated Word Recognition Method Based on Vector Taylor Series

A technology of vector Taylor series and recognition method is applied in the field of isolated word recognition based on multi-environment model based on vector Taylor series, which can solve the problems of unsatisfactory recognition performance of speech recognition system, reduce false recognition rate and improve performance. Effect

Active Publication Date: 2017-09-01
SOUTHEAST UNIV
View PDF2 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Although the speech recognition system has a high recognition performance in the laboratory, in the actual environment, due to the variability of speech and the interference of environmental noise, the recognition performance of the speech recognition system is not satisfactory.

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 Multi-environment Model Isolated Word Recognition Method Based on Vector Taylor Series
  • A Multi-environment Model Isolated Word Recognition Method Based on Vector Taylor Series
  • A Multi-environment Model Isolated Word Recognition Method Based on Vector Taylor Series

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0025] The present invention will be further described below in conjunction with the accompanying drawings.

[0026] The present invention will be further described below.

[0027] The present invention provides a method for identifying isolated words based on a vector Taylor series (VTS: Vector Taylor Series) multi-environment model (VTSME: VTS-based Multi-Environment), which mainly includes two stages: a training stage and a recognition stage.

[0028] In the model training stage, set the basic environment including SNR of 0dB, 5dB, 10dB, 15dB, 20dB and pure environment, according to the set SNR parameters, weight the white noise and load it into the pure training speech to obtain different Noisy training speech in SNR environment. In each basic environment, the MFCC parameters of the noisy training speech are extracted, and the noisy GMM (including the pure GMM model) and the noisy HMM model (including the pure HMM model) are trained and generated respectively. These noisy...

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 multi-environment model isolated word recognition method based on vector Taylor series, through the model training stage: set the basic environment signal-to-noise ratio, use noisy training speech to train and generate noisy GMM model and noisy HMM model respectively; through isolated word recognition stage, according to the noisy GMM model obtained in the training stage, first select the SNR environment that best matches the current test environment; secondly, based on the vector Taylor series and the pure GMM model in the pure environment, estimate the mean and variance of the noise in the test speech , and according to the minimum mean square error criterion, the test speech feature parameters are mapped to the noisy speech feature parameters in the most matching SNR environment; finally, the noisy HMM model in the best matching SNR environment is selected, and the mapped containing The noise feature parameters are matched with the corresponding noisy HMM model to obtain the final recognition result. The misrecognition rate of the invention is greatly lower than that of the existing vector Taylor series.

Description

technical field [0001] The invention relates to the field of speech recognition, in particular to a vector Taylor series-based multi-environment model isolated word recognition method. Background technique [0002] Speech recognition is a wide range of subjects, including: signal processing, mathematical statistics, pattern recognition, acoustics and phonetics, artificial intelligence and so on. With the rapid development of modern science and technology, speech recognition technology has gradually stepped out of the laboratory and entered people's daily life. Although the speech recognition system has a high recognition performance in the laboratory, in the actual environment, due to the variability of speech and the interference of environmental noise, the recognition performance of the speech recognition system is not satisfactory. Therefore, it is of great practical significance to study the speech recognition technology and improve the robustness of the speech recognit...

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): G10L15/14
Inventor 周琳束佳明吕勇吴镇扬
Owner SOUTHEAST UNIV