An architecture for 1-bit localized compressive sensing with applications to EEG

Compressed sensing exploits special signal features to extract its information content with a smaller amount of samples with respect to acquisition based on Nyquist theorem. While many theoretical results have proved the capabilities of this new paradigm, hardware implementations are still far from being practical. Here, we present a new architecture of analog to information converter that produces 1-bit compressive measurements. The performance of the architecture can be boosted if the signal to acquire features, beyond the classically required sparsity, also some sort of localization of its energy. The effectiveness of the architecture and of its enhancement is shown in the measurement of EEG, that presents a non-uniform spectral profile.