This study assesses the perceived urgency of warning sounds according to different acoustic parameters. After a definition of a set of acoustic parameters (pulse duration, interpulse interval, frequency, number of bursts, bursts interval, timbre), two methods were investigated and compared in order to understand and optimize the perceived urgency of alert sounds. The first method uses a classical D-optimal Design of experiments, the second method is based on an Interactive Genetic Algorithm (IGA) for the optimization of the fitness of a population of sounds. With these two methods, two within-subjects perceptual tests were carried out during a first session with a panel of 30 participants. To compare the results, a second session was organized with the same participants in order to assess the performances of the optimal sounds generated by both methods. Individual sounds and also sounds at the group level were compared, and the possible interactions between the different acoustics parameters of the alert sounds were investigated. Results shows that the IGA method can be an interesting alternative to classical DOE to help the design of sounds