战场多目标混合声信号源数盲估计算法研究

Source Number Blind Estimation Algorithm of Battlefield Multiple Target Acoustic Signal Mixture

为了适应战场的实际应用环境,解决传统的MDL准则信源数估计算法在有色噪声条件下失效的问题,提出了基于MDL函数值比值的信源数目估计算法,但此算法在低信噪比下信源数的检测概率较低。针对这一问题,用样本特征值平方和与样本特征值之和的比来代替MDL函数中的样本特征值算术平均的方法来改进MDL准则函数,从而得到基于改进的MDL函数值比值算法。该算法可以有效提高信源数所对应的MDL函数值与非信源数对应的MDL函数值之间大小的差距,从而进一步提高低信噪比条件下MDL比值法的估计性能。通过仿真试验和实测数据的验证试验,表明所提出的两种算法是有效的和可行的。

In order to adapt for the battlefield environment and solve an invalidation problem of estimation of source number using traditional minimum description length （MDL） criterion for colored noise, a novel algorithm based on the ratio of MDL criterion function was proposed, but its detection probability is lower for lower signal to-noise ratio. For this question, the arithmetic mean of the sample eigenvalues was replace with the ratio of sum of sample eigenvalue square to that of the sample eigenvalues to modify MDL criterion function. The modified algorithm can increase the difference between MDL function value associated with the source number and that with non-source number and improve estimating performance of the ratio of MDL criterion function under low signal-to-noise ratio. The pro- posed algorithms were evaluated in various conditions by using simulated and measured experimental data. The results of the experimental demonstrate the validity and feasibility of the proposed algorithms.