Artificial Neural Network Application In Sense Signal's Pattern Recognition

As the rapid development of neural network theory, the study of neural network had nearly reached every field of our life. Neural Network has much advantage such as an adaptive capability, good fault-tolerance and adapt to VLSI system and so on. This thesis focused on the study of artificial neural network application in Sensor signal's pattern recognition, Used the LAB VIEW and MATLAB language to design a hybrid programming, to get the amplitude and RMS, and use the neural network analysis method to make a pattern recognition of the sensing signal, the application program can easily realized the function to detect the output signal in the distributed optical fiber sensing system and analysis the types of the external disturbance signals in the real-time way, so it can use the distributed optical fiber sensing system to make a real-time detect and judge of the monitoring objects. This significance of the article is to provide a new way about the pattern recognition of the fiber sensor signal, look forward to bring some help on the exploring the new and more effective training method of the neural network's pattern recognition.The main research focuses on the MATLAB neural network's design, and the transplant of the pattern recognition program from MATLAB to LABVIEW.The main research contents are listed as following.(1) Study the structure and function of the Mach-Zehnder optical interferometer sensing system and artificial neural network, and to do the analysis of the Mach-Zehnder optical interferometer system's demodulation and demodulation error in further.(2) Building a distributed optical fiber sensing system to get the sample data, and using LabVIEW to make the interception analysis and processing of the data. Select the amplitude and RMS of the segment data as the feature of sensing signals.(3) Designed for four improved BP neural network in MATLAB program, and use the obtained characteristics of the sample data to train these neural network, make a comparison of these training results, select the best training method of BP network. To get the best effect of the sensor signal's pattern recognition.(4) Select the LMBP neural network which based on numerical optimization theory algorithms and probabilistic neural network and the corresponding training method to get the pattern recognition of the sensor signal. Design the corresponding MATLAB programs and use the obtained characteristics of the sample data to train the neural network. So that it can get the capacity to accurately identify the different patterns of external interference signals.(5) Make a further study of hybrid programming methods between MATLAB and LAB VIEW, design and use the MATLAB script node of LAB VIEW in the hybrid programming, from this way can realize the function to design the friendly user interface by using LABVIEW software, at the same time it can improve the algorithm's process efficiency, and the algorithm can work in real-time.