Mobile Activity Recognition through Training Labels with Inaccurate Activity Segments (Best paper candidate)

Reviewed, Featured
Takamichi Toda, Sozo Inoue, Naonori Ueda,
International Conference on Mobile and Ubiquitous Systems: Computing, Networking and Services (MobiQuitous) (Acceptance Rate: 30%)
(Not Available)
(Not Available)
In this paper, we propose an approach to improve mobile activity recognition given training dataset with inaccurate segments, in which the beginning and the ending timestamps of homogeneous and continuous activities have inaccurate boundaries by human errors.
In the proposed approach, we A) convert the training dataset to multi-label samples, B) train the dataset with multi-label EM learning algorithm, and C) apply a segmentation method using not only the estimated labels, but also the original segment information.
We evaluate the proposed approach for 3 datasets including simulation data and real activity data, 2 machine-learning algorithms, and various inaccuracies, and show that the proposed approach outperformed the naive methods for the questions of Q1) fixing the segments of the training data and Q2) improving the recognition accuracy by cross validation.

Data Files