Research And FPGA Implementation Of Signal Recognition Based On Convolutional Neural Network

Introducing the machine learning into power line communication,convolutional neural network is used to realize OFDM signal recognition of power line transmission characteristic.Under the three platforms of Matlab,Caffe and Tensorflow,from two-dimensional picture signal to one-dimensional signal data,the proposed method is modeled and simulated to realize classification and recognition.Firstly,nine kinds of OFDM signals with different transmission parameters are obtained by using the Matlab to simulate the low-voltage power line communication channel model.The Alexnet toolkit in the Matlab is used and adjusted as a learning network.The saved signal recognition network model is used for the final signal recognition,realizing batch classification and accurate recognition of each signal.Secondly,the characteristics,framework and construction pro cess of the Caffe platform are briefly introduced.In the aspects of data format conversion,creating neural network and training and testing signal,the specific process of realizing signal classification is described,and the final recognition result s are given.On this basis,the Open VINO toolkit proposed by Intel is introduced to further verify the realization of signal recognition on the Caffe platform,and the recognition results and reasoning time are given.Next,on the Tensorflow platform,denoising auto-encoder is introduced to realize the additive noise denoising of OFDM signal.On this basis,the signal recognition is realized,which not only verifies the denoising effect but also realizes the signal classification.The most important thing is th at when two-dimensional signal picture is used as input,the RGB processing is introduced for interference after denoising,which achieves a good denoising effect.Combined with the practical application,the denoising auto-encoder is improved to realize the denoising of one-dimensional signal data,and the neural network is used to complete the further classification and recognition.Finally,on the basis of one-dimensional signal,combined with the signal recognition network model trained on the Tensorflow platform,the Open CL signal recognition based on FPGA is completed.It is verified that the FPGA implementation of signal recognition not only ensures the recognition accuracy,but also embodies its acceleration function.