时幅连续信号数值反卷积方法及其应用

为简单有效地恢复脉冲辐射场诊断信号，研究提出一种数值反卷积方法。在一定条件下对输出序列和响应序列进行抽样，构造“超定”线性卷积方程组，并在一定物理约束下利用最小二乘法求解方程组，得到较低采样频率下输入信号序列的反卷积恢复。在Matlab中实现了该算法并进行了模拟。结果表明，当输出序列和响应序列所含加性白噪声小于5%时，无需任何预处理，反卷积结果与原输入序列符合很好。该方法适用于时幅连续信号处理，能较好地解决反卷积的病态问题，使用方便，且不需序列重组，结果直观、不需延迟修正。该方法在脉冲辐射场诊断中已得到成功应用。

Numerical Deconvolution Method for Continuous Signals Processing and Its Application

A numerical deconvolution method is presented to recover signals of diagnoses in pulsed radiation fields simply and effectively. The vector of deconvolution recovery is obtained at a lower sampling frequency from an ‘over determined’ convolution linear equations set which is constructed by sampling output vectors and response vectors at a certain condition and is solved by least squares method at some physical constraints. The method is realized and simulated in Matlab. The simulated results show that the deconvolution solutions agree well with the input vectors without any additional process when the errors associated with the output vectors and the response vectors are less than 5%. The method is suitable for continuous signals processing and very accuracy results can be obtained in ill- posed problems solving. The reconstruction of the output vectors and the response vectors and the correction of the results are not needed by this method, which can be realized and operated conveniently. As an example, a successful application of the method in pulsed radiation field diagnosis was reported.