Method and device for parameterized expression, encrypted transmission and reconstruction of voice signals

Abstract

The invention provides a voice signal parameterized expression, encrypted transmission and reconstruction method and device based on data driving. According to the method, signals mixed with frequencycomponents are decomposed into low-frequency components and band-pass components based on frequency band division, wherein the frequency components of the low-frequency components and the band-pass components are relatively simple; a Magic function is used as a kernel function, each frequency component can be expressed as a linear combination of the kernel function, and a corresponding coefficient becomes a parameter group for uniquely determining each frequency component; the Magic function shows a frequency window characteristic in a frequency domain, the main lobe of the Singer function shows the characteristic of sharp reduction similar to a wavelet function in a time domain, the method provided by the invention not only has very high resolution in a frequency domain, but also has locality in a time domain, can reflect the change of each frequency component at a certain time point, is suitable for processing non-stationary signals similar to voice, realizes parameterized expression of voice signals, and provides conditions for encryption transmission. By using parameters of a voice signal, the voice signal can be reconstructed and restored uniquely.

Description

technical field [0001] The invention relates to the field of speech signal processing and analysis, in particular to a method and device for parameterized expression, encrypted transmission and reconstruction of speech signals. Background technique [0002] At present, voice signals are generally stored in the format of WAV files. If you want to compress voice files without reducing the voice quality, the best way is to decompose the voice signal in the time-frequency domain, so as to express it in a parametric form. For speech signals, if the parametric expression of speech signals can be realized, not only can the storage volume of speech signals be compressed to the greatest extent, but also encrypted transmission and fidelity reconstruction can be easily realized. [0003] Fourier series and Fourier transform are the earliest tools for signal processing and data analysis. Using this tool, the time domain signal can be transformed into the frequency domain, and the freque...

AI Technical Summary

Problems solved by technology

[0004] Wavelet transform can transform non-stationary signals into the time-frequency domain. It is appropriate to use wavelet transform to study the frequency components of speech signals, but the correlation between points in the time-frequency domain of wavelet transform is so large that wavelet transform The data redundancy of the system is very large, which cannot meet the requirements for compressing and storing the voice signal. Furthermore, the inverse wavelet transform is not a 1-1 mapping, which leads to difficulties in the process of encrypting and reconstructing the voice signal.

[0005] In recent years, a method combining Empirical Mode Decomposition (EMD) and Hilbert Transform (HHT) has emerged to analyze non-stationary signals. This method can accurately obtain the corresponding relationship between the time and frequency of the signal. Many application fields have achieved relatively good results, but the empirical mode decomposition is not based on a clear mathematical model, so the parametric expression of the speech signal cannot be realized

Images