【摘要】 针对电动轿车低速行驶时车外声音过小而难以引起车辆和行人的警觉问题,本文中提出一种警示声的设计方法。首先,采集传统内燃机汽车车外声,并通过采用语音合成技术改变其声音信号特定频段内的幅值,以获得大量的声音信号样本;接着,以声品质概念为基础,进行主观评价试验,并计算出各声音信号的客观心理声学参数,选出对应车速区间内评分值最好的声音样本作为警示声;最后,采用模拟退火算法和遗传算法来优化BP神经网络,并以主客观评价数据为基础,建立SA/GA-BP神经网络的电动轿车警示声声品质主观评价客观量化模型,以6个客观参量作为输入,主观评价值作为输出。结果表明,该模型具有很快的收敛速度和很高的预测精度。更多还原

【Abstract】 Aiming at the problem that electric vehicle in low-speed driving is too quiet to arouse the vigilance of surrounding vehicles and pedestrians,the design method of a warning sound is proposed in this paper. Firstly,the external noise of traditional vehicle with internal combustion engine is collected,and by using speech synthesis technology to change the amplitudes of its sound signals in specific frequency band,a large number of sound signal samples are obtained. Then,based on the concept of sound quality,a subjective evaluation test is carried out,and the objective psychoacoustic parameters of each sound signal are calculated,from which the sound samples with the best score in the corresponding speed interval are selected as warning sounds. Finally,BP neural network is optimized by simulated annealing( SA) algorithm and genetic algorithm( GA),and based on subjective and objective evaluation data,a model objectively quantifying subjective evaluations of the quality of warning sounds for electric vehicle is established by SA/GA-BP neural network with 6 objective parameters as inputs and subjective evaluation value as output. The results show that the model set up has a fast convergence speed and a high prediction accuracy.更多还原