Voice data classification method based on improved capsule network

Abstract

The present invention discloses a voice data classification method based on an improved capsule network. The method comprises the steps of: in a training phase, constructing a coder of a capsule network to perform coding of initial voice phoneme data to obtain a primary capsule; constructing a dynamic routing structure of the capsule network to transmit the information in the primary capsule to ahigh-class capsule; taking a softmax excitation value of the length of each high-class capsule to represent the probability of the initial voice phoneme data belonging to a corresponding type; constructing a decoder of the capsule network to perform decoding reconstruction of the high-class capsule corresponding to real phoneme symbols; performing optimization of parameters of the capsule networkbased on a total loss function; and in a test phase, inputting the initial voice phoneme data to the coder of the capsule network to determine the type of data to be tested. The voice data classification method based on the improved capsule network can perform accurate identification of the phoneme symbols corresponding to the time sequence signals of the voice phonemes so as to solve the technical problems that the directly extracted feature accuracy is low, the voice data classification effect is poor and the over-fitting is performed according to the theoretical knowledge.

Description

technical field [0001] The invention belongs to the technical field of classification processing and deep learning, and in particular relates to a speech data classification method based on an improved capsule network. Background technique [0002] Speech data is an important processing content of modern information data. Each frame of speech data can be described by characteristic parameters, such as formant related parameters, that is, the formant frequency (first dimension) and bandwidth (second dimension) of a frame of speech data. ), energy spectrum tilt (third dimension), etc., the above are multi-dimensional features directly extracted based on the experience of researchers and theoretical knowledge. However, such work is very computationally intensive and requires a lot of experimentation and innovation. The deep learning method that has emerged in recent years integrates feature extraction and feature classification. It has very powerful feature self-organization a...

AI Technical Summary

Problems solved by technology

[0003] The purpose of the present invention is to propose a voice data classification method based on the improved capsule network, to realize the accurate recognition of the sound velocity symbol corresponding to the time sequence signal of the voice phoneme, and to solve the problem that the voice features directly extracted according to the theoretical knowledge in the prior art are accurate. The technical problems of low precision, poor voice data classification and over-fitting

Images