Research For Feature Extraction And Pattern Recognition Of Typical Signals

Research object of this paper are communication signals and biomedical signals,the main content of the study for communication signal are feature extraction and modulation recognition,and for biomedical signal are feature extraction and cluster analysis,based on sensors.For communication signals,features are extracted,using the combination of traditional signal processing theory and artificial neural network method,and using the combined method of time domain and frequency domain based on the theory of decision tree to identify seven modulation signals.With summarizing the characteristics of various modulation method,the author breaks the traditional feature extraction methods,strengthens the features,and extracts features for the second time.This method effectively improves the recognition rate and algorithm robustness.During the process of implementation,the author broke through traditional single feature extraction methods,by summing up and studying different modulation mode.according to these characteristics,and on the basis of feature extraction for the first time,I do secondary feature extraction,namely after extracting the feature to extract feature vector again,effectively improving the recognition rate and recognition algorithm of stability.In the study of the single signal modulation mode recognition,on the basis of the separation and identification of mixed signal,I used the band-pass filter to separate mixed signal into separate signals,and then according to the identification of a single signal.Currently it can only achieve a mixture of at most two signal separation and identification.In the process of the communication signal modulation mode identification,it needs to eliminate interference and noise,the author studied some common jamming signal recognition,including gaussian signal,narrow pulse signal and radiate transformer power supply of the signal.These kinds of signals and on the principle of signal modulation signal is different,only through the combined method of time domain and frequency domain can be identified.This article communication signals modulation mode identification in the process of receiving equipment for the oscilloscope,its advantage is strong commonality,easy to instrument communication and interaction with laptop computers.According to the signals’ characteristics of envelope,signals can be divided into constant envelope signals and non-constant envelope signals,respectively using power ratio method and(SVD)singular value decomposition method.In this paper,the author completes six kinds of signals’ SNR estimation by simulation.In terms of biomedical signals,only based on smartphones,data mining and analysis techniques are used to measure many kinds of indicators of medicine anytime and anywhere,without any medical equipment.At present,I developed a mobile phone APP,which can measure heart rate.And the measurement on blood pressure still stays in the simulation stage,the topic need more time for further research.With the increasing focus on health and the popularity of the smartphones,I believe that such a study is meaningful and there will be more people to develop in this field in the future.