Improvement of the lighting fixtures based indoor localization method CEPHEID

This paper deals with the author’s indoor positioning method named as CEPHEID (Ceiling Embedded PHoto-Echo ID), which is proposed recently. It uses flickering of lighting fixtures as an environmental fingerprint. It is characterized by employing deep neural network including 1D CNN as discriminator. However, there has been a question that whether such a costly computation as DNN is indeed necessary or not. In this paper, the author firstly introduces original CEPHEID and shows its high performances through two experiments. Then, a discussion of using SVM as its classifier aiming to reduce computation cost is described. To evaluate SVM classifiers, the author performs the third experiment by using the same data. As a result, SVM classifiers shows poor performance than DNN one. Consequently, DNN can be regarded as a not-too-much or an acceptable technique for CEPHEID.