Study On Feature Extraction And Classification Of Doppler Ultrasonic Signal

Ultrasound is a sound whose frequency tops out 20 kHz. It has not been heard by human. Along with the Austrian scientists discovered that the Doppler effect is phenomenon of acoustic, ultrasound is widely applied, particularly in the medical field. In the medical images, ultrasound equipment are usually ultrasonic ranging in frequency from 2M to 20M k, which are currently widely used in clinical diagnosis and are demanded greatly. On the base of the principle of Doppler ultrasound, Doppler ultrasound in medical technology has been used to measure the diagnosis of human blood and heart and other organs of the velocity of blood parameters, which have been become a clinical diagnosis of the important means currently.At present, to a certain extent, Doppler spectrum image reflect the physiological and pathological vascular status. It has played an important role in the clinical diagnosis of the disease. The Doppler spectrum image analysis can only play the role of qualitative, lack of quantitative of the diagnosis of diseases associated with error. Especially in recent years, in order to further explore Doppler ultrasound technology in the clinical diagnosis of the role and value and go deep into study Doppler ultrasound signal contained information by analyzing various parameters of Doppler spectrum,which have not been previously recognized features extracted through various methods. It is essential to reduce diagnostic error, reflect disease of quantitative.Feature extraction and classification of Doppler transcranial ultrasound detector based on Audio signals is studied in this paper by the rapid development of wavelet theory and the theory of SVM research tools in recent years. Wavelet transform having characteristics of multi-resolution, by coarse-to-fine observed signal, time and frequency of the local and so on, can approximation to a function or signal by a family of function, can characterize the original signal by extracting the main features information of the signal , Audio signals of transcranial Doppler of ultrasound detector is studied in this paper. Hanning windowed short-time Fourier transform is used to get the spectra of ultrasound Doppler signals, and then percentage method is used to the maximum frequency curve, on the basis of the maximum frequency curve, the characteristics of audio is acquired by modulus maximum of wavelet transform. The SVM which have been put forward first by the former Soviet Union professor's Vapnik is based on New machine learning of statistical learning theory,Different from traditional study method, the method is built on the VC-dimensional statistical learning and SRM. According to the limited sample information is between the complexity and the study ability of model which looks for the best compromised of complexity and study ability. At last according to feature information of audio signals, classification is realized by SVM theory. The simulation results show that using the method introduced in the paper can the diagnosis of vascular disease, provide prevention to assistance, provide the objective of Accordings in the staging of patients.