Barraza Jose F, Grzywacz Norberto M
Departamento de Luminotecnia, Luz y Visión, Universidad Nacional de Tucumán, and Consejo Nacional de Investigaciones Científicas y Técnicas, Av. Independencia 1800, T4002BLR San Miguel de Tucumán, Tucumán, Argentina.
Vision Res. 2008 Oct;48(23-24):2485-91. doi: 10.1016/j.visres.2008.08.011. Epub 2008 Sep 24.
If the purpose of adaptation is to fit sensory systems to different environments, it may implement an optimization of the system. What the optimum is depends on the statistics of these environments. Therefore, the system should update its parameters as the environment changes. A Kalman-filtering strategy performs such an update optimally by combining current estimations of the environment with those from the past. We investigate whether the visual system uses such a strategy for speed adaptation. We performed a matching-speed experiment to evaluate the time course of adaptation to an abrupt velocity change. Experimental results are in agreement with Kalman-modeling predictions for speed adaptation. When subjects adapt to a low speed and it suddenly increases, the time course of adaptation presents two phases, namely, a rapid decrease of perceived speed followed by a slower phase. In contrast, when speed changes from fast to slow, adaptation presents a single phase. In the Kalman-model simulations, this asymmetry is due to the prevalence of low speeds in natural images. However, this asymmetry disappears both experimentally and in simulations when the adapting stimulus is noisy. In both transitions, adaptation now occurs in a single phase. Finally, the model also predicts the change in sensitivity to speed discrimination produced by the adaptation.
如果适应的目的是使感觉系统适应不同的环境,那么它可能会实现系统的优化。最优状态取决于这些环境的统计数据。因此,系统应随着环境变化更新其参数。卡尔曼滤波策略通过将当前对环境的估计与过去的估计相结合,以最优方式执行这种更新。我们研究视觉系统是否使用这种策略进行速度适应。我们进行了一个匹配速度实验,以评估对突然速度变化的适应时间进程。实验结果与速度适应的卡尔曼模型预测一致。当受试者适应低速并突然增加时,适应时间进程呈现两个阶段,即感知速度的快速下降随后是较慢阶段。相反,当速度从快变为慢时,适应呈现一个阶段。在卡尔曼模型模拟中,这种不对称性是由于自然图像中低速的普遍性。然而,当适应刺激有噪声时,这种不对称性在实验和模拟中都消失了。在两种转变中,适应现在都在一个阶段发生。最后,该模型还预测了适应引起的速度辨别敏感性的变化。