In consideration of the overcrowded spectrum resource, it is of great importance to study the combination of the cognitive radio (CR) and ultra wideband impulse radio (UWB-IR). In the cognitive UWB scene, the UWB pulse is supposed to constantly change its spectrum according to the state of the primary users to mitigate the potential interference to these vulnerable wireless systems. Meanwhile, the frequency utilization of the UWB pulse still eagerly requires improvement to guarantee their transmission reliability. In this paper, we propose a novel UWB pulse based on the neural network, and its simple implementation has also been presented. The frequency utilization of our designed UWB pulse can surpass any other existing UWB signals when no primary user is working around. Based on spectrum sensing information, our designed UWB pulse can adaptively alter its output spectrum to effectively avoid interference to the primary users. We also demonstrate that the proposed UWB pulse is much more suitable for the FCC emission limit than the ECC emission limit.