Anti-alias Demodulation Of Analog-digital Hybrid AM Broadcasting System

With the continuous development of modern communication technologies,traditional analog amplitude modulation(AM)broadcasting has been far from meeting people's needs.Digitalization has become an inevitable trend in the development of the world's broadcasting and television.However,spectrum resources are invaluable in digital communications.Therefore,in the process of digitizing analog broadcasts,higher spectrum utilization must be ensured.The M-aryPosition Phase Shift Keying(MPPSK)-based analog-digital hybrid AM broadcasting system,which maintains the original double sideband AM scheme and uses the MPPSK digital carrier instead of a sinusoidal carrier to carry the amplitude modulation of analog broadcast signals.This not only achieves the simulcast of analog audio and data,but also greatly improves the spectrum utilization.However,the shaping filtering to meet the Federal Communications Commission's requirements for AM transmit signal power spectrum has introduced strong ISI,which seriously affects the demodulation performance of digital signals,makes the non-encoding demodulation error rate only tend to 1% magnitude.In such a case,this thesis proposes a detection method,which applies the deep learning(DL)algorithm for classification problems to demodulate MPPSK signals.Much research works have been done about the digital signal demodulation of analog-digital hybrid AM broadcasting systems with severe ISI.Firstly,this paper puts great emphasis on the research of the analog-digital hybrid AM broadcasting systems,including the MPPSK transmission system,the system framework,system modulation and demodulation principle and working process.The digital modulation,audio modulation,complex modulation,shaping filtering,audio demodulation,and data demodulation are introduced in detail,and the demodulation performance of analog audio and digital signals are given at last.Secondly,this paper introduces the typical DL algorithms,including Deep Neural Network(DNN),Convolutional Neural Network(CNN)and Stacked Auto-Encoder(SAE),and analyzes the unique advantages of DL network in dealing with classification problems,proposes the idea of MPPSK classification detection based on DL network,and discusses its feasibility and advantages.Then,MPPSK demodulators based on DL detection network are proposed.Simulations on demodulation performance of the three MPPSK demodulators,DL-DNN,DL-CNN and DL-SAE,shows that the MPPSK demodulation error rate within these three DL detection networks can reach the order of 10e-3,which is an order of magnitude better than those the traditional detection algorithms are,and the DL-SAE algorithm works best.Finally,the paper proposes a multi-symbol united-decision demodulator based on DL detection network,using the multi-symbol united-decision method to solve the problem of waveform extension and distortion due to ISI,improve the demodulation performance of MPPSK signals.By optimizing themulti-symbol united-decision scheme,an optimal MPPSK demodulation scheme suitable for the studied system is designed,and the demodulation error rate of the MPPSK modulated signal can be reduced to the order of 10e-4,which is two orders of magnitude better than those the traditional MPPSK detectors are,and one order of magnitude better than the mono-symbol detection scheme.