Today most people, especially the elderly in urban areas, spend more than 80% of their time in indoor environments, such as at home, the workplace, public and private places. Indoor human tracking is a basic task thats of high value in variety of applications, such as security system, person surveillance and human-centered ambient intelligence. Recently, optical fiber sensor has been explored for device-free human tracking, which offers a promising alternative to video or wearable sensor based paradigm, because of its advantages of data-efficient, low-computational-cost and non-intrusive in indoor environments. The challenges in building fiber-optic sensors based human tracking system are to accurately and quickly estimate human position, as well as provide acceptable efficiency performance. Motivated by these challenges, we have built a multiplexing sensing based tracking system using optical fiber sensors. To derive an accurate and efficient tracking scheme, we show that the position information of human body can be formulated as bipedal movement on the constructed grids of an indoor structured environment. The properties of spatio-temporal correlations and sparsity of the bipedal movements is then incorporated into the multiplexing sensing mechanism for efficiency enhancement. In addition, a look-up-table is applied to map the multiple measurement vector into the target position. Finally, a lab-scale prototype system is developed for indoor human tracking. The experimental results demonstrate the effectiveness of the proposed method in human tracking.