Mobile trace inference based on tensor voting

Erte Pan, Miao Pan, Zhu Han, Vernaldo Wright

Abstract: As the growth of wireless network, tremendous interests have been focused on statistically tracking the user equipment as well as the performance evaluation of motion tracking. In this paper, we tackle the problem of inferring human mobility trace under the circumstance that the recorded location information exhibits missing data. Based on the tensor voting theory, we propose an efficient sparse tensor voting algorithm and a specified implementation scheme. The model is constructed based on the geometric connections between the input signals and encodes the structure information in the tensor matrix. Thus, the computation is carried out in the form of matrix, which reduces the computation load since most the calculation involves only with matrix addition and multiplication. The proposed method is applied to real human mobility trace. The results show that our proposed approach effectively recovers human mobility trace from the incomplete data input.