研究一类具有时变时滞及参数不确性的Cohen-Grossberg神经网络的鲁棒稳定性问题.应用划分时滞区间的思想构造了一个新的Lyapunov泛函,并以线性矩阵不等式的形式给出了平衡点全局鲁棒稳定性判据,新判据放松了时变时滞变化率必须小于1的限制.仿真结果进一步证明了所得结论的有效性. A robust stability problem for a class of Cohen-Grossberg neural networks with time-varying delays and parametric uncertainties is studied. A new Lyapunov functional is constructed using the idea of ​​partitioning dead-time intervals, and is given as a linear matrix inequality The criterion of global robust stability of the equilibrium point is obtained, and the new criterion relaxes the restriction that the rate of change of time-varying delay must be less than 1. The simulation results further prove the validity of the conclusion.