PESA: Probabilistic Efficient Storage Algorithm for Time-Domain Spectrum Measurements

Wireless communication is an essential part of daily life for users globally with applications in medical devices, cellular phones, Internet of Things nodes, and others. Accordingly, there is a need to understand the patterns and properties of radio-frequency spectrum use by acquiring accurate spectrum utilization measurements. However, the massive storage volume needed to execute spectrum surveys-especially when a fast sampling rate is used-is an impeding factor in terms of cost and ease-of-access. In this paper, a probabilistic efficient storage algorithm (PESA) is proposed to facilitate high-accuracy, time-domain spectrum surveys conducted at a fast sample acquisition rate to detect sporadic spectrum occupancy patterns that could be in the order of microseconds. PESA divides the dynamic range of monitoring equipment into bins-each represented by one component of a Gaussian mixture model (GMM). Windows of activity and inactivity in the measurements are established by comparing with a threshold, and then indicators to the GMM component that best describes a window are recorded, hence reducing the required storage volume. Results demonstrate that {\approx}99 % reduction in the storage volume is achievable while maintaining an accurate estimation of channel utilization (CU) and activity/inactivity periods. Furthermore, a LabVIEW implementation of PESA on a hardware platform was executed and used to survey Wi-Fi channel 1 in a healthcare environment for seven consecutive hours. Although more than 25 billion samples were observed, the resulting data only occupied 96.28 MB.