Voice signal feature learning method based on first derivative of Mel-spectrogram

A first-order derivative and voice signal technology, applied in voice analysis, instruments, etc., to achieve the effects of fast speed and scalability, good discrimination, and less training time

Active Publication Date: 2018-11-06
浙江中点人工智能科技有限公司
View PDF11 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

But this kind of diagnosis depends on the doctor's personal senses and the valuable experience accumulated in the long-term practice of medicine, and this experience cannot be copied

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
  • Voice signal feature learning method based on first derivative of Mel-spectrogram
  • Voice signal feature learning method based on first derivative of Mel-spectrogram

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0026] The application will be described in further detail below in conjunction with the accompanying drawings. It is necessary to point out that the following specific embodiments are only used to further illustrate the application, and cannot be interpreted as limiting the protection scope of the application. The above application content makes some non-essential improvements and adjustments to this application.

[0027] combine figure 1 , figure 2 As shown, the speech signal feature learning method based on the mel spectrum first derivative of the present invention comprises the steps:

[0028] Step 1. Input disease speech samples and healthy speech samples;

[0029] Step 2. Framing all samples, detecting speech endpoints, extracting the first derivative of Mel spectrum MFCC with respect to time DMS (first Derivative of Mel-Spectrogram), and using matrix A for each sample i express;

[0030] The analysis of MFCC is based on the auditory principle of the human ear, whic...

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 provides a voice signal feature learning method based on a first derivative of Mel-spectrogram (DMS). The method comprises inputting a disease voice sample and a healthy voice sample based on data driving; splitting all samples into frames to extract the DMS versus time, determining the training sets and the test sets of the disease sample and healthy sample by a cross-validation method; training dictionaries for healthy voice and pathological voice separately by using a clustering algorithm, subjecting the DMS of each sample in the two training sets and the two test sets to linear coding and to pooling by using a minimum pooling method to obtain the final features. The supervised method makes full use of tag information, and the learned features have better discriminating power.

Description

technical field [0001] The invention relates to the field of artificial intelligence speech recognition, in particular to a speech signal feature learning method based on the first derivative of Mel spectrum. Background technique [0002] The method of diagnosing diseases by sound has received widespread attention in recent years because of its advantages of simplicity, convenience, speed, and no need to damage the patient's body and invasive examination. Studies have shown that speech signals contain rich biomedical information. For example, speech can become very soft, and eventually develop into a monotonous, non-fluctuating voice, and it can be judged that an individual may suffer from Parkinson's disease. When an individual has thyroid disease, it can lead to hormonal imbalances that can even lead to paralysis or paralysis of the vocal cords, which can make the voice muffled and sometimes even whisper-like. By extracting and analyzing the biological information feature...

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): G10L25/18G10L25/27G10L25/48G10L25/66G10L17/04
CPCG10L17/04G10L25/18G10L25/27G10L25/48G10L25/66
Inventor 朱成华卢光明武克斌张大鹏钟德才
Owner 浙江中点人工智能科技有限公司
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