In order to describe the rapid change of nuclear pulse signal, the fine time-frequency resolution method was adopted.By this time-frequency method, nuclear pulse signals were decomposed layer by layer, and the optimal sub-band spaces with minimum Shannon entropy were selected adaptively, then the important coefficients of each space were extracted as characteristic parameters to describe the signals.Results show that, with very few characteristic parameters, for example account for 6%of raw data size, the error of pulse ampli-tude and time constant was less than 0.4%, the important information of waveform and its singular points was kept intact.This method is important to subsequent treatment processes such as shaping pulse and gaining pa-rameters.%针对核脉冲信号具有随时间快速及奇异变化特点，提出了一种基于时频分辨的精细描述方法。对核脉冲信号进行多层时频分解，自适应选择香农熵意义下的最优子频带空间，并提取各空间重要系数作为描述信号的特征参数。结果表明，用来描述信号的特征参数仅占原始数据量的6％便可将脉冲幅值及其时间常数误差控制在0．4％以内，波形及其奇异点重要信息保持完好，对于后续核脉冲成形及参数获取等数字处理过程的研究具有重要意义。
展开▼
