A Deep Learning Method for Speech Gender Recognition

Abstract

The invention discloses a voice gender recognition deep learning method, comprising the following steps: a. collecting voice signals; b. detecting the endpoints of the collected voice signals and segmenting the human voice voice signal segments; c. The voice signal segment is divided into frames and multi-resolution cochlear voice features are extracted for each frame; d. Input the voice features of each frame into the pre-trained deep learning model of voice gender recognition for classification; e. Determine the output voice gender Voting statistics, according to the final output of the speaker's gender of the voice signal segment, the present invention uses multi-resolution cochlear voice features, which are more in line with the voice feature parameters of human auditory perception analysis, and uses capsule networks as the voice gender recognition acoustic reasoning model , can adapt to the speech environment with low signal-to-noise ratio, and has a higher recognition rate than traditional methods.

Description

technical field [0001] The invention relates to a deep learning method, and more specifically, relates to a deep learning method for speech gender recognition, which belongs to the technical field of gender recognition. Background technique [0002] Speaker recognition is a hotspot in the field of identity authentication and artificial intelligence. Solving the problem of speaker recognition has important theoretical value and far-reaching practical significance. Gender recognition can be applied to occasions where the gender of boys and girls needs to be verified, such as knowing the gender of the person being verified from the first digit of the ID card number, and verifying whether the speaker’s gender is consistent with the gender contained in the ID card number. It can be applied to the front-end of speech recognition. When the gender of the speaker is recognized, the gender-dependent speech model (Gender-Dependent Model) is used for recognition to improve the speech re...

AI Technical Summary

Problems solved by technology

The current speech gender recognition system usually consists of three parts, which are speech signal preprocessing, feature extraction and classification. Feature extraction is the most important part, and its quality directly affects the recognition results. The speech gender proposed by previous researchers Most of the features are based on the prosody features and sound quality features of speech, which are all artificially designed features. The robustness of the system is not high, and it is easily affected by the environment.

Images