利用截断奇异值分解法解冲激响应逆卷积

Deconvolution of Impulse Response by the Truncted Signular Value Decomposition

在目标识别领域中,目标的冲激响应占有十分重要的地位,然而冲激响应必须通过解电磁系统逆卷积求得。本文直接从时域的观点出发,采用截断的奇异值分解法解这类病态问题,去掉了引起病态的较小奇异值,使这类逆卷积问题趋于良态,并对解球目标的冲激响应进行了数值仿真。结果表明,使用截断奇异值分解法比传统的共轭梯度法得到的解改善了约10dB。

The impluse response of the target plays a very important part in the target identification. However, the impluse response of an electronmagnatic system can be obtained through solving the deconvolution problem. In this paper,the ill-posed problem is solved directly in the time domain by truncted singular value decomposition (TSVD) method. The less singular values which cause the problem ill-posed are discarded,so that the deconvolution problem is changed into a illess-posed problem. The impluse response of a sphere is used to make some simulation. The results show that the TSVD algorithm can improve the result of the deconvolution by about 10 dB(SNR) compared with the conventional conjugated gradient method (CGM).