Speech depression state recognition method based on feature selection and transfer learning

A technology of transfer learning and feature selection, applied in speech recognition, speech analysis, instruments, etc., can solve the problems of widening the distribution of speech signal features of the subjects, unsatisfied assumptions, and increasing the difficulty of model recognition, etc., to achieve low Model complexity, optimal recognition accuracy, and the effect of improving model efficiency

CN113555004APending Publication Date: 2021-10-26FUDAN UNIV
16 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Publication Date
2021-10-26

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention provides a speech depression state recognition method based on feature selection and transfer learning, and provides the speech depression state recognition method fusing Lasso and a transfer learning method CORAL in order to solve the two problems that the feature dimension is higher, and the feature distribution is affected by the tested individual differences except for the depression level during modeling based on speech. The method has the advantages that the redundant information in Lasso filtering features is reserved, the effective features are reserved, and the recognition precision is further improved on the basis of improving the model efficiency; 2, on the premise that the depression label information is not leaked by the transfer learning method CORAL, the feature distribution of a training set and a test set is drawn close, and the influence of other factors except the depression level on the feature distribution is reduced; the combination of the two methods can further improve the accuracy and stability of depression screening.
Need to check novelty before this filing date? Find Prior Art

Description

Technical field

[0001] The invention belongs to the field of speech signal processing, and specifically relates to a speech depression state recognition method based on feature selection and transfer learning. Background technique

[0002] Depression is a typical and common mental illness around the world, covering all ages. The current clinical diagnosis method of depression relies on the clinical experience of doctors and related scales filled out by patients. The whole process takes a long time, and the diagnosis The process is inefficient. Voice, as an important external expression of emotion, has become a key direction for researchers to implement automated depression identification methods due to its unique advantages such as few restrictions on use, low equipment cost, no contact, and non-invasive and convenient collection methods.

[0003] Currently, there are no clear specific features supported by theoretical background for depression identification. The feature d...

Examples

Embodiment 1

[0020] figure 1 It is a flowchart of a speech suppression state recognition method based on feature selection and transfer learning according to an embodiment of the present invention.

[0021] like figure 1 As shown, the speech suppression state recognition method based on feature selection and transfer learning in the embodiment of the present invention includes the following steps:

[0022] Step S1, voice information collection, use recording equipment to collect voice, design questions of different types of speech tasks, the subjects answer according to the prompts on the screen, use the recording equipment to collect the complete speaking process of the subjects, and record it as a wav file, This file is the speech sample.

[0023] Step S2, voice signal preprocessing, preprocessing the collected voice samples, manual screening to exclude obvious noise segments, such as coughing, sounds of things falling, and high-pass filtering, down-sampling, silent segment detection a...

Embodiment 2

[0064] As mentioned above, Embodiment 1 provides a speech depression state recognition method based on feature selection and transfer learning, which mainly includes steps S1 to S6. In actual application, each step of the method in Embodiment 1 can be configured as a corresponding computer module, that is, a speech collection part, a preprocessing part, a feature extraction part, a feature processing part, a transfer learning part, and a classification part, and these parts form a A device for classifying and identifying speech depression states, thereby also providing a speech depression state recognition device based on feature selection and transfer learning.

[0065] figure 2 It is a schematic diagram of a speech depression state recognition device based on feature selection and transfer learning according to an embodiment of the present invention.

[0066] like figure 2 As shown, the speech depression state recognition device based on feature selection and transfer le...