j whose j-th action characteristic time sequence xj is the maximum likelihood, and for, when the k-th action characteristic time sequence is xk, calculating action characteristic generation probability p(xk|θj) with respect to arbitrary j, k; a voting rate calculation part for acquiring the p(xk|θj) with respect to the arbitrary j, k and action characteristic data d, and for calculating a voting rate p(c'|θj) with respect to the arbitrary j; and a voting rate storage part for storing the voting rate p(c'|θj).;COPYRIGHT: (C)2012,JPO&INPIT"/>
INFORMATION NECESSITY/NON-NECESSITY LEARNING ESTIMATION DEVICE, INFORMATION NECESSITY/NON-NECESSITY LEARNING ESTIMATION METHOD AND PROGRAM
PROBLEM TO BE SOLVED: To accurately estimate probability that a user needs information.;SOLUTION: The information necessity/non-necessity learning estimation device includes: a necessity/non-necessity extraction part for acquiring the action of a user, and for, when information is requested, outputting "necessity", and for, otherwise, outputting "non-necessity" as necessity/non-necessity information c; an action characteristic extraction part for extracting an action characteristic time sequence x from a moving image obtained by photographing a user; an action characteristic storage part for storing the combination of the necessity/non-necessity information c with the action characteristic time sequence x just before the acquisition of the necessity/non-necessity information c as action characteristic data d; an action characteristic generation probability calculation part for acquiring the action characteristic time sequence x, and for calculating a parameter θj whose j-th action characteristic time sequence xj is the maximum likelihood, and for, when the k-th action characteristic time sequence is xk, calculating action characteristic generation probability p(xk|θj) with respect to arbitrary j, k; a voting rate calculation part for acquiring the p(xk|θj) with respect to the arbitrary j, k and action characteristic data d, and for calculating a voting rate p(c'|θj) with respect to the arbitrary j; and a voting rate storage part for storing the voting rate p(c'|θj).;COPYRIGHT: (C)2012,JPO&INPIT
展开▼
机译：解决的问题：准确地估计用户需要信息的可能性。解决方案：信息必要性/非必要性学习估计装置包括：用于获取用户动作的必要性/非必要性提取部分，用于何时请求信息，输出“必要”，否则，输出“不必要”作为必要/不必要信息c;动作特征提取部分，用于从通过拍摄用户而获得的运动图像中提取动作特征时间序列x;动作特征存储部，在即将取得必要/非必要信息c之前，将必要/非必要信息c与动作特征时序x的组合存储为动作特征数据d。动作特征产生概率计算部分，用于获取动作特征时序x，并计算第j个动作特征时序x j 为的参数θ j 最大似然，并且，当第k个动作特征时间序列为x k 时，计算动作特征产生概率p（x k |θ j < / Sub>）关于任意j，k;投票率计算部分，用于针对任意j，k和动作特征数据d获取p（x k |θ j ），并计算投票相对于任意j的比率p（c'|＆theta; j ）; COPYRIGHT：（C）2012，JPO＆INPIT;以及投票率存储部分，用于存储投票率p（c'| theta; j ）。
展开▼
