Simmons J A
Cognition. 1989 Nov;33(1-2):155-99. doi: 10.1016/0010-0277(89)90009-7.
Echolocating bats perceive objects as acoustic images derived from echoes of the ultrasonic sounds they emit. They can detect, track, identify, and intercept flying insects using sonar. Many species, such as the big brown bat, Eptesicus fuscus, emit frequency-modulated sonar sounds and perceive the distance to targets, or target range, from the delay of echoes. For Eptesicus, a point-target's image has a sharpness along the range axis that is determined by the acuity of echo-delay perception, which is about 10 ns under favorable conditions. The image as a whole has a fine range structure that corresponds to the cross-correlation function between emissions and echoes. A complex target- which has reflecting points, called "glints", located at slightly different distances and reflects echoes containing overlapping components with slightly different delays--is perceived in terms of its range profile. The separation of the glints along the range dimension is encoded by the shape of the echo spectrum created by interference between overlapping echo components. However, Eptesicus transforms the echo spectrum back into an estimate of the original delay separation of echo components. The bat thus converts spectral cues into elements of an image expressed in terms of range. The absolute range of the nearest glint is encoded by the arrival time of the earliest echo component, and the spectrally encoded range separation of additional glints is referred to this time-encoded reference range for the image as a whole. Each individual glint is represented by a cross-correlation function for its own echo component, the nearest of which is computed directly from arrival-time measurements while further ones are computed by transformation of the echo spectrum. The bat then sums the cross-correlation functions for multiple glints to form the entire image of the complex target. Range and shape are two distinct features of targets that are separately encoded by the bat's auditory system, but the bat perceives unitary images that require fusion of these features to create a synthetic psychological dimension of range. The bat's use of cross-correlation-like images reveals neural computations that achieve fusion of stimulus features and offers an example of high-level operations involved in the formation of perceptual "wholes".
使用回声定位的蝙蝠将物体视为由它们发出的超声波回声所形成的声学图像。它们能够利用声呐探测、追踪、识别并拦截飞行中的昆虫。许多物种，比如大棕蝠（Eptesicus fuscus），会发出调频声呐信号，并根据回声的延迟来感知到目标的距离，即目标范围。对于大棕蝠来说，点目标的图像在距离轴上具有锐度，这取决于回声延迟感知的敏锐度，在有利条件下约为10纳秒。整个图像具有精细的距离结构，这与发射信号和回声之间的互相关函数相对应。一个复杂目标——其具有位于稍有不同距离处的反射点，即“闪烁点”，并且反射的回声包含具有稍有不同延迟的重叠成分——是根据其距离剖面图来感知的。沿着距离维度闪烁点的分离是由重叠回声成分之间的干涉所产生的回声频谱形状编码的。然而，大棕蝠会将回声频谱转换回对回声成分原始延迟分离的估计。因此，蝙蝠将频谱线索转换为以距离表示的图像元素。最近闪烁点的绝对距离由最早回声成分的到达时间编码，而其他闪烁点经频谱编码的距离分离则相对于整个图像的这个由时间编码的参考距离。每个单独的闪烁点由其自身回声成分的互相关函数表示，其中最近的闪烁点直接根据到达时间测量值计算得出，而更远的闪烁点则通过回声频谱的变换计算得出。然后，蝙蝠将多个闪烁点的互相关函数相加，以形成复杂目标的完整图像。距离和形状是目标的两个不同特征，由蝙蝠的听觉系统分别编码，但蝙蝠感知到的是单一图像，这需要融合这些特征以创建一个综合的距离心理维度。蝙蝠对类似互相关图像的使用揭示了实现刺激特征融合的神经计算，并提供了一个参与感知“整体”形成的高级操作的例子。
