To Recognize Signal And Forecast Using Time Series And Support Vector Machine

Signal recognition problem is a very important issue in the field of information transmission now. Research on acoustic signal recognition has made great development, especially in the military field. Countries have invested a lot of money and researchers for in-depth study. For example, in military sonar recognition technology. Sonar is main technique to monitor underwater by States Navy, used for underwater target detection, classification, location and tracking, underwater communications and navigation, protect the ships, anti-submarine aircraft and anti-submarine helicopters. Sonar technology has become a national focus on the development of underwater technology, objects, along with the development of mathematical theory and signal processing techniques, sonar signal identification method become diversify, identification of technical performance greatly improved, identification methods and models varied, but the main advantage of this in several ways:neural network, wavelet transform, support vector machine, power spectral and the combination of these methods.This paper use time series and support vector classification machine to identify the signal. Some practical problems due to the occurrence and development of information on randomness, and with the passage of time have some statistical law, in this case, difficult to apply analytical methods to determine the general description of the process, and time series analysis is used statistics and information data processing techniques, to explore its own rules and proved their identity, is a very effective tool to solve practical problems, and the method is more mature, and its parameters estimated and tests were also easier to operate. Support vector machine is a new data mining technology, is a machine learning new tools by optimization methods to solve the problem. It is very successful in dealing with classification, discriminant analysis, pattern recognition and regression problems, and many other issues, and it can be used to predict and comprehensive evaluation and other fields. This paper uses time series model, equaling measure alternate time series model and ARTAFIT model to deal with the acoustic wave signal. Then it can get the parameter set. For the different acoustic wave signal, it can get the different parameter set. It use support vector machine to recognize acoustic wave signal through parameter set. When the parameter set is very big, it can be first used the method of principal composition analysis to lower the dimension, then it will be used the method of support vector machine to recognize the acoustic wave signal. At the end, it gives the example. The result of the application is ideal.Prediction for the future is one of the important uses of time series model. Time series forecasting method is through the preparation and analysis of time series, according to the time sequence reflected in the development process, direction and trends, carried out by analogy or extension in order to predict the next period of time or after a period of years may reach levels. Because a lot of data in the reality are nonlinear, many of the traditional method for the effect of non-linear prediction problem are not very good. Therefore, this paper use of time series and support vector regression machine to prediction. First of all, in order to extract more useful information, this paper use of support vector regression machine to the original fitting and forecasting time series, using support vector regression machine can be a high degree of fit the characteristics of nonlinear time series, to effectively improve the prediction accuracy. Then get a residual sequence, while the residual sequence is often still contain some useful information, and sometimes there will be for predicting the results of greater impact. To extract the useful information of the residuals, this paper use of time series model fitting and prediction residual sequence. And, ultimately, support vector regression and time series forecasting models combining, we can get more accurate forecasts.