Indoor White Space Measurement at Different Cities
In order to study the characteristics of the indoor white spaces, we carried out a long-term indoor white space measurement at two cities of China: Shanghai and Nanjing.
We first implemented an energy detector based on the GUN Radio platform and USRP N210 devices. More than 22 USRPs were utilized to measure the RSSIs of TV spectrum in different buildings. The measurement lasted for more than one month. We collected roughly 3400000 samples and characterized the spatio-temporal-spectral correlations among indoor TV spectrum based on the collected samples.
FIWEX: Cost-efficient Indoor White Space Exploration
FIWEX is a cost-efficient indoor white space exploration system.
A compressive sensing based data reconstruction algorithm is utilized to revocer the indoor white space availability map from the partial sensing results. A k-medoids clustering based sesnor deployment algorithm is utilized to determine the proper locations of sensors.
FIWEX identifies 47.8% more indoor white spaces with 38.4% less false alarms compared to the state-of-the-art system.
TIME: Training-free Indoor White Space Exploration
TIME is a training-free indoor white space exploration system. Existing systems rely on the specific spatio-spectral correlations of indoor white spaces from the training process, which needs a lot of devices, energy, and man-power. TIME instead utilizes the general redundancy among white spaces, and applies Relevance Vector Machine (RVM) and Bayesian compressive sensing to construct the indoor white space availability map.
TIME achieves competitive performance with the training-based systems
FRISE: Fine-grained Indoor White Space Exploration
FRISE is a fine-grained indoor white space exploratin system which could accurately identify the whit space availabilities at arbitrary indoor locations.
FRISE characterizes the spatio-temporal-spectral correlations of indoor white space by multiplying a Gaussian kernel, a period kernel, and a semi-positive matrix together, and utilizes multitask Gaussian process model to achieve the fine-grained indoor white space estimation.
FRISE determines the positions of the candidate locations based on a mutual infomation maximization algorithm.
Mobile Sensor Based Indoor White Space Exploration
In this project, we have designed and implemented a mobile spectrum sensing platform which contains an USRP N210, an omni antenna, an Arduino-WiFi based mobile device and an Intel NUC. We tend to implement an indoor white space exploration system based on the mobile spectrum sesning platform, and sharply reduce the cost and complexity.
We have designed two mechanisms: one is based on compressive sening and has no requirement for the route of the mobile sesnor; the other one is based on multitask Gaussian process and the route of the mobile sensors are real-timely calculated by maximizing the mutual information.
This project is in preparation for MobiCom 2017.