Compression of Channelized Wideband Spectrum Samples

Abstract

As spectrum becomes more congested, spectrum monitoring becomes increasingly necessary. Spectrum monitoring generates large amounts of IQ data, often placing strains on storage capacity. For these systems, data compression offers to reduce the volume of the incoming data, allowing for increased recording time or bandwidth for analysis. This thesis introduces a lossy compression technique for wideband IQ data, which uses channelization, bounding box detection, and quantization via the golden quantizer for two-term Gaussian mixture distributions. The results show improved MSE performance and comparable CR when compared to using an adaptive vector quantizer.