Research And Application Of Human Voice Signal Based On Signal Adaptive Decomposition Technology

With the gradual improvement of people's living standards,the topic of health has also received more and more attention.In fact,many patients in our country die of chronic diseases such as cardiovascular and cerebrovascular diseases,diabetes,and cancer every year.An important reason for this is that the early diagnosis of the disease is wrong,so that patients miss the best treatment time,resulting in increased mortality.At present,there is still a huge shortage of professional medical talents with rich clinical diagnosis experience in China.Due to the development of artificial intelligence technology and a new generation of information technology,the early disease auxiliary diagnosis system based on artificial intelligence technology has gradually become a research hot spot.This study combines the machine learning technology in artificial intelligence with the adaptive signal technology of the signal,and applies it to the human voice signal to further establish an early auxiliary diagnosis system for diseases.Human voice signals are not only affected by different vocal organs during the production process,but also interfered by external stimuli during the operation of the human body.It can be seen that the human voice signal is a nonlinear and non-stationary signal.Traditional signal processing techniques such as Fourier transform,filter techniques and wavelet analysis are not suitable for this.In this regard,according to the characteristics of human voice signals,this thesis conducts theoretical and methodological research,and the main achievements are as follows:(1)Aiming at the problem that patients need long-term mouth breathing after nasal surgery,a method of mouth breathing training before nasal surgery using breathing sounds is proposed.The method uses empirical mode decomposition to analyze and extract features of breath sound segments,and uses cluster analysis method to pre-screen the data of breath sound segments,and finally uses random forest for classification.(2)Aiming at the problem of early diagnosis of vocal cord diseases,a classification method of vocal cord pathology based on empirical mode decomposition and electroglottograms was proposed.The method uses empirical mode decomposition to decompose the electroglottograms,further uses the cluster analysis method to group the obtained intrinsic mode functions,then performs feature extraction,and finally uses random forest for classification.(3)Aiming at the problem of early diagnosis of Alzheimer's disease,a speech detection method for Alzheimer's disease based on variational mode decomposition and random forest is proposed.This method uses variational mode decomposition to decompose the speech signal,further proposes a variational mode function selection method based on random forest algorithm,then extracts the features of the selected variational mode function,and finally uses random forest for classification.All the proposed algorithms have been theoretically analyzed and verified by simulation.The results show that the model established in this thesis and the method proposed in this thesis can effectively classify the respiratory part or perform early diagnosis of related diseases of human voice signals.