Abstract
While classical compressive sensing aims to reduce the number of measurements with respect to Nyquist-based sampling methods, it is usually important to consider the total number of bits needed to represent this small amount of measurements in order to maintain the same signal quality. In this paper we study an architecture for signal acquisition that produces a stream of 1-bit measurements, and we show that a proper exploitation of its features, as well as of the sparse representation of the input signal can result in an extremely high reconstruction accuracy with a total number of bits much smaller than classical sampling.