Research On Detection Algorithm Of Weak Telemetry Signal

A necessary and basic step in the communication process is signal detection,whether the presence of a signal involves the communication system for subsequent demodulation and other processing.In telemetry communication system,the signal-to-noise ratio of the received signal is low due to the long communication distance,and due to the high-speed movement of the aircraft,the received signal is affected by the doppler frequency shift.Therefore,this paper mainly studies signal detection algorithms based on spectral features and combined with classification techniques in machine learning in low signal-to-noise ratio and high dynamic environments.This paper firstly introduces the PCM-CDMA-BPSK signal model commonly used in the telemetry communication system.Because it is a spread spectrum signal,it has strong antiinterference ability.At the same time,the received signal model after the signal passes through the channel is also introduced.Then,the principles of several commonly used signal detection algorithms are introduced,and the probability density distribution of detection statistics is deduced from the existing algorithms,and the theoretical expressions corresponding to detection probability and false alarm probability are derived.Finally,the advantages and disadvantages of commonly used detection algorithms are analyzed.Based on the advantages and disadvantages of existing algorithms and the characteristics of signal and noise,this paper proposes two detection algorithms for measurement and control signals in low signal-to-noise ratio and high dynamic environments.The first algorithm is a detection algorithm based on spectral features—dynamic spectral detection algorithm.This algorithm is mainly based on the continuity of the two-channel signals of the shift sampling and the non-continuity of noise.According to the time-shift property of Fourier transform.The two channels of signals after delayed sampling have a certain linear relationship in the frequency domain,and the two channels of noise after delayed sampling are random,so after certain processing of the received signal according to this relationship,when there is a target signal,the detection statistic gradually increases with the accumulation of time,and when there is no target signal,the detection statistic is almost unchanged with the accumulation of time.Therefore,as long as the accumulated points are large enough,the presence or absence of the target signal can be distinguished according to the size of the detection statistic.And through the simulation results,it is found that this algorithm has certain robustness to Doppler frequency shift.Then,wavelet decomposition is introduced for the problem of low signal-to-noise ratio.By decomposing the signal into multiple layers,each layer contains different signal components,and some noise components are suppressed to obtain a denoised signal.The simulation results show that the detection performance has been further improved under low signal-to-noise ratio.In addition,the principle of determining the threshold value in the wavelet decomposition is also improved,and change the original method of using the same threshold for each layer to a smaller threshold as the number of decomposition layers increases.This rule is mainly based on the fact that the noise attenuation increases with the increase of the number of decomposition layers.The simulation results show that the presence or absence of the target signal can be reliably detected at a lower signal-to-noise ratio after improving the threshold determination principle.The second algorithm is a detection algorithm based on the classification technology in machine learning.Compared with the traditional signal detection algorithm combined with the classification technology in machine learning,the signal needs to be preprocessed at the receiver to extract signal features.This paper proposes a detection algorithm that does not require signal preprocessing at the receiver.The simulation results show that this algorithm can reliably detect the target signal under the condition of low signal-to-noise ratio,and also reduces the complexity.In addition,the traditional signal detection algorithm combining cyclic spectrum and machine learning classification technology does not fully utilize the spectral information.The simulation results show that the performance of the improved algorithm has been further improved compared with the original algorithm.At the same time,due to the combination of the random forest classification algorithm with strong classification characteristics,it has better detection performance under low signal-to-noise ratio than the algorithm not combined with machine learning classification technology.