Haptic Recognition System Based on Sensory Integration and Active Perception
Haptic Recognition System Based on Sensory Integration and Active Perception
复制标题
基于感觉统合和主动感知的触觉识别系统
DOI：
10.9746/sicetr1965.31.1217
复制
发表时间：
1995
期刊：
影响因子：
--
通讯作者：
Y. Sakaguchi
中科院分区：
文献类型：
--
作者：
Y. Sakaguchi
当人类通过触摸感知物体时，根据识别的过程，以适当的方式，例如摩擦，推动和拾取，将各种感觉信息与观察物体的表面相结合。这一事实表明，感觉整合和主动感知在触觉识别过程中起着至关重要的作用。根据触觉学的这些特点，作者构造了一个触觉识别系统，它以类似于人的方式描述触觉。该系统配备了几个传感器设备，包括振动传感器，摩擦传感器和热传感器，可以用几个力和速度值推动和摩擦物体的表面。该方法根据识别过程的不同，选择合适的传感器和测量条件，迭代地集成来自这些传感器的信息。感觉整合和主动感知的算法分别由贝叶斯推理和基于信息准则的迭代实验设计实现。实验结果表明，该系统可以区分触感的细微差异：可以几乎完美地区分16种纸张和布料。证明了系统根据识别过程选择合适的传感器，即主动感知算法比随机感知算法用更少的观测值实现了更好的识别精度。此外，它是实验表明，在系统中使用的特性，以及对应于那些人类利用触觉识别。这些结果表明，所构建的系统是一个忠实的模型的人的触觉机制。
When perceiving an object by touch, human beings integrate various sensory information with observing the object's surface in appropriate manners, such as rubbing, pushing and picking, according to the proceeding of the recognition. This fact suggests that sensory integration and active perception play essential roles in the haptic recognition process. In light of such characteristics of haptics, the author constructed a haptic recognition system which descriminated feel of touch in a similar manner to human's. The system is equipped with several sensor devices, including a vibration sensor, a friction sensor and a thermal sensor, and can push and rub the object's surface with several values of force and speed. It integrates information from these sensors iteratively with selecting an appropriate sensor and a measurement condition according to the proceeding of the recognition. The algorithms of sensory integration and of active perception are realized by Bayes inference and by an iterative experimental design based on an information criterion, respectively. The experimental result shows that the system can discriminate subtle difference in feel of touch : It can discriminate 16 kinds of paper and cloth almost perfectly. It is also proved that the system selects appropriate sensors according to the proceeding of the recogntion, that is, the active perception algorithm realizes good recognition accuracy by fewer observations than the random observation algorithm. In addition, it is shown experimentally that the characteristics used in the system well correspond to those human beings utilize in haptic recognition. These results suggest that the constructed system is a faithful model of the human haptic mechanism.