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

Learning device, voice interval detector, and method for detecting voice activity

A sound interval and learning device technology, which is applied in neural learning methods, speech analysis, biological neural network models, etc., can solve the problems of low detection accuracy of sound intervals and the inability to properly distinguish noise and sound, and achieve the effect of improving detection accuracy

Active Publication Date: 2020-10-16
MITSUBISHI ELECTRIC CORP
View PDF9 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0010] The method described in Patent Document 1 has a problem that, in the case of an unknown noise environment not assumed in the learning data used for learning the HMM of noise, noise and sound cannot be properly distinguished, and the noise interval may be misjudged as sound. Interval, the detection accuracy of the sound interval is low

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
  • Learning device, voice interval detector, and method for detecting voice activity
  • Learning device, voice interval detector, and method for detecting voice activity
  • Learning device, voice interval detector, and method for detecting voice activity

Examples

Experimental program
Comparison scheme
Effect test

Embodiment approach 1

[0031] figure 1 It is a block diagram showing the configuration of the speech interval detection system 1 including the learning device 2 and the speech interval detection device 3 according to Embodiment 1 of the present invention. The learning device 2 generates a synthetic neural network (hereinafter referred to as synthetic NN) b by inputting learning data a, and learns a Gaussian mixture model of noise and sound (hereinafter referred to as noise and sound GMM) c. The voice interval detection device 3 detects the voice interval of the input signal based on the synthesized NN b, the noise and the voice GMM c, and the noise Gaussian mixture model (hereinafter referred to as noise GMM) d, and outputs the voice interval detection result.

[0032] Learning data a is data including spectral feature quantities of noise data and voice data. The spectral feature quantity is, for example, 1-dimensional to 12-dimensional vector data of Mel-frequency cepstral coefficients (hereinafte...

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 present invention corrects a level of sound serving as a discriminant measure to determine the boundary between noise and voice using a Gaussian mixture model for noise learned during a time interval when an input signal is noise, and detects a voice interval on the basis of the corrected level of sound.

Description

technical field [0001] The present invention relates to a learning device for sound interval detection of an input signal, a sound interval detection device and a sound interval detection method. Background technique [0002] In the voice recognition process, pattern recognition is performed on the voice interval detected from the input signal to obtain the recognition result. Therefore, if there is an error in the detection of the voice interval, the recognition accuracy of the voice recognition process will be greatly reduced. In the detection of the audio interval, there is a method of detecting an interval in which the power of an input signal is equal to or greater than a threshold value as an audio interval. This method is effective in a relatively small and stable environment with background noise. [0003] On the other hand, in the input of inspection results in the maintenance work of plant equipment, or in the operation support of various factory automation equipm...

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): G10L25/78
CPCG10L25/84G10L25/30G06N3/084G10L2025/783G06N3/045G06N7/01
Inventor 花泽利行
Owner MITSUBISHI ELECTRIC CORP