Sound Collision Classification Based On Support Vector Machine Wheat

Food is the basic substances for human, and can directly affect the survival and development of human. But during the stored, the grain is often damaged by insects and damaging by mice and mold prone, and the damage will reduce the grain production and cause the decline of the quality issues, and metamorphic food will cause harm to human health. Therefore, it is particularly important to find effective detection and prevention work. In recent years, the method of sound signal detection has becoming a hot research, because it is light, simple, fast and cheap. In the field of acoustic detection, the scholars at home and abroad have done a series of research on the technology of impact sound, for this method is convenience and environmentally sustainable. And this method has been gradually applied to detection of insect pest and sorting of grain.This article is takeing three wheat impact sounds of undamaged kernels, insect damaged kernels and moldy kernels as study object and proposing pattern recognition method that based on support vector machine to classify those impact sounds. The impact sounds is analysised and processed by the signal methods of bispectrum, wavelet packet and EEMD, and extract effective features of acoustics. Then the features were classified in support vector machine, and the recognition rate is very good. This paper provides a new method and scientific basis for wheat kernels classification and grain detecting.This paper mainly includes the following aspects:(1) Introduce research background and significance of the field, and reviews the development present situation of impact sounds detection technology.(2) Introduce the support vector machine, detailing tell the basic theory, principle, classification and application of the method.(3) Using a new approach that combined bispectrum method and support vector machine to classify and research the impact acoustic signals of wheat kernels. The impact acoustic signals were processed by bispectrum estimation. After analsied, based on the difference of singal, features in bispectrum and diagonal slices spectrum were extracted. Then the features were classified in support vector machine. The recognition accuracy rates in classification of undamaged kernel, insect damaged kernel and moldy kernel were above84%, and the accury rate of insect damaged kernels is as high as94%. The experimental results show that the feature of wheat impact sounds is very clear in bispecturm, the classify result of suppot vector machine is very good.(4) The method of combining wavelet packet and support vector machine was used to classify and analyse the three types of impact acoustic signals of wheat kernels. First introduces the principle of wavelet packet and wavelet packet decomposition reconstruction theory, etc. And select the best wavelet packet basis function to decompose the signals. There is a big difference of each node of three types of impact acoustic signals. The characteristic features including node energy, singularity value, power spectral entropy and spectral entropy arm of node envelope signals were extracted, and the features were classified in support vector machine.There is a high recognition accuracy rate in classification of three types of wheat kernels. The experimental result shows the identification of the method of the grains is effective, and each type of wheat impact signals features are much different, and this research provides a new method for wheat kernels sorting.(5) Using the method combined EEMD and support vector machine to classify and study the impact acoustic signals of wheat kenerls. First this paper introduced the decomposition of the EEMD method principle and the related development application. After several tests, the EEMD parameter value was determined. After EEMD decomposed, we found that there is a certain differences between signals in different frequenc, and signals waveform is adifferent in different frequency domain, the information of impact signals is mainly concentrated in the high frequency part. To calculate the IMF component energy, IMF component kurtosis and Renyi entropy of high frequency part as feature vector, then the features were classified in support vector machine. The recognition accuracy rates are very good, it provides a new method and scientific basis for wheat kernels classification.