基于卷积神经网络的OFDM-UWB信道环境识别

超宽带（UWB）无线通信技术被广泛应用于室内定位领域，其能识别出多径信道中的非视距信道，有助于去除影响信号的非理想因素，提升定位精度。基于OFDM方案的UWB系统，提出一种利用卷积神经网络对信道估计出的信道冲激响应时频图像进行训练，从而识别出信道环境的方法，将信道识别问题转化为图像识别问题，同时分析时频处理参数对识别结果的影响。仿真结果表明，该方法的识别率随通信系统比特信噪比（EbN0）的增加而提升，当EbN0增加至20 dB时稳定在90%，与传统基于支持向量机的信道识别方法相比获得了10%的性能提升。

OFDM-UWB Channel Environment Recognition Based on Convolutional Neural Network

Widely used in the field of indoor positioning,the Ultra-WideBand(UWB) wireless communication technology can identify the Non-Line of Sight(NLOS) channel in the multi-path channel,helping in removing non-ideal factors and improving positioning accuracy.Based on the OFDM-UWB system,this paper proposes a LOS/NLOS channel recognition method.The method employs a Convolutional Neural Network(CNN) to train the estimated time-frequency images of channel impulse responses,transforming the channel recognition problem into an image recognition problem.In addition,the influence of time-frequency processing parameters on the recognition results is analyzed.The simulation results show that the recognition rate of this method increases with the bit signal-to-noise ratio(EbN0) of the communication system,and is stable at 90% when EbN0 increases to 20 dB.At the same time,the method provides 10% performance improvement compared with the traditional channel recognition method based on Support Vector Machine(SVM).