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

Automatic speech recognition channel normalization

A normalization, speech technology, applied in speech recognition, speech analysis, instruments, etc., can solve problems such as uncertainty limitation, and achieve the effect of improving robustness, reducing the volume of speech, and reducing system delay

Inactive Publication Date: 2008-07-23
VOICE SIGNAL TECH
View PDF0 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, attempts to use fixed ratios have been limited by the uncertainties involved in distinguishing between speech and non-speech segments

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
  • Automatic speech recognition channel normalization

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0027] A processing system for automatic speech recognition channel normalization includes offline processing and online processing to generate normalization parameters. The system is configured to utilize observations about properties of communication channels. For example, the following observations about speakers and parts of the communication channel - including room, microphone and ambient noise - can be made:

[0028] • The long-term spectrum of a speaker can be characterized mainly by two parameters: the overall loudness and the spectral tilt which describes the overall slope of the spectrum. The spectral tilt is a direct result of the ratio of the time the glottis remains open versus closed during each pitch period. The spectral tilt is typically -12dB / octave, although the ratio varies slightly across speakers and their vocal effort (normal, shouting). In the cepstral domain, overall loudness is captured by 0-order cepstral coefficients, while spectral tilt is captur...

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

Statistics are measured from an initial portion of a speech utterance. Feature normalization parameters are estimated based on the measured statistics and a statistically derived mapping relating measured statistics and feature normalization parameters.

Description

[0001] Cross References to Related Applications [0002] This application claims priority to US Provisional Application Serial No. 60 / 535,863, filed January 12,2004. technical field [0003] The present invention relates to channel normalization for automatic speech recognition. Background technique [0004] The recognition performance (eg, accuracy) of an automatic speech recognition system can be adversely affected by the variability of the communication channel. Some causes of variability are due to: the speaker (e.g., vocal tract geometry, glottal excitation), the transmission channel (e.g., variable position and orientation to the microphone, room acoustics, ambient noise), and Use a microphone with different characteristics. In order to reduce the impact of the communication channel on the recognition performance, many schemes have been proposed. One such technique normalizes the identified feature vectors of the cepstral coefficients such that each feature dimensio...

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 Applications(China)
IPC IPC(8): G10L19/14G10L15/00
CPCG10L15/02G10L25/24G10L15/20
Inventor 伊戈·兹洛卡尼克劳伦斯·S·吉利克乔丹·科亨
Owner VOICE SIGNAL TECH