Perception of Simulated Propeller Cavitation
提出一个随机模型描述人对复杂多维声音（如被动声纳信号）的分类过程，通过模拟螺旋桨空化声实验验证模型，并讨论其对声纳操作员训练的价值。
A stochastic model is proposed to describe the perceptual processes that underlie the classification of complex multidimensional sounds, such as passive sonar signatures. The model is based on four assumptions: (1) complex sounds may be represented psychologically as points in a multidimensional perceptual space, (2) categories are represented theoretically as distributions in this space centered at ideal category members or prototype, (3) the category distributions determine the conditional stimulus probabilities and have multivariate Gaussian form, and (4) individual dimensions in the perceptual space may be weighted differentially to reflect a listener's attentional processes. To evaluate the model, two groups were tested in an eight-category classification task involving 16 simulated propeller cavitation sounds. A perceptual structure was determined for the sounds in a preliminary non-metric multidimensional scaling study. The two groups received category assignments that stressed different properties of the cavitation sounds. An observed confusion matrix was derived by fitting the model to the observed data. Overall, comparison of the obtained and theoretical data indicated that the model provides a reasonable description of how listeners classify complex simulated sonar sounds. The assumption that listeners can selectively and independently adjust the relative importance of the perceptual dimensions was also supported. The utility of the proposed model as a tool for the development of preprocessing aids and training programs for sonar operators is discussed.