基于时频分析的噪声去除

经典的数字信道化思想的实施都是建立在一组硬件滤波器构成的并行多路,虽然能实时处理,但随着处理数据带宽的增大和子信道频率分辨率的增加,设备开销将线性增长,整体可靠性也必然下降.基于信号的检测理论,在利用时频分析工具得到的时频图上进行信号检测,提出了检测去噪法.针对宽带数据背景噪声的特点,提出了噪声基底的3种估计算法:去大留小法、对数统计法和频移均值法,并线面结合对噪声基底进行平滑,实现了天电干扰的去除.通过实际接收到的某短波信号检验,证明了算法的有效性.

The De-noising Based on Time Frequency Analysis

The idea of actualization of classical figure channel is on the basis of collateral multi-path,which is com- posed of a set of hardware filters.Though it can process data real-timely,the equipment spending will linearly increase while the holistic reliability will inevitably goes down,as the processing data bandwidth enlarges and the sub-channel frequency reso- lution increases.The article brings forward the method of detection noise removing,based on the theory of signal detection and signal detecting on the time frequency chart gotten by time frequency analysis tools.In allusion to the feature of broadband data background noise,three estimating algorithms of background noise are advanced:the discarding the big and keeping the small method,the logarithm statistical method and the frequency shift averaging method.Besides,the background noise is smoothed with the line and surface bonding method,and the atmospheric interference is actually wiped off.Through the test of certain short -wave signals in actual reception,the availability of the algorithms is proved.