中文摘要：

通过对安装于机器人手爪上的阵列触觉传感器所采集到的触觉图像的分析,提出了用Hopfield神经网络实现触觉图像恢复的技术。Hopfield神经网络把触觉图像的每一个像素作为一个处理单元（神经元）,像素之间的关系即神经元之间的权值作为储存单元。首先利用Hebb规则计算权值矩阵,用以存储所有样本的标准模式,然后利用网络的联想记忆能力恢复被抓物体的触觉图像。实验结果表明,该方法能达到很好的触觉图像恢复和识别效果。

英文摘要：

The technique of tactile image restoration using Hopfield neural network is presented through analyz- ing tactile image collected with robotic paw tactile sensor array. Hopfield neural network regards every pixel of the tactile image as a processing unit （neuron） and the relation among pixels that is the weight of the neural network as a memory cell. Firstly, the weight matrix that is used to store the standard mode of all samples is calculated using Hebb rule. Then the tactile image is restored using the associative memory of Hopfield neural network. Experiment shows that the proposed approach is effective in tactile image restoration and recognition.