Circuits, Systems, and Algorithms for Low-power Signal Processing in IoT Nodes Implementation

Overview

In the IoT paradigm the low-power signal processing, either analog of digital, is a key-enabling technology. Many unconventional processing techniques, either based on a statistical analysis or not, have been introduced in the effort of being able to complete a task with the lowest possible energy.
A first example is given by the Compressed Sensing, an acquisition technique which relies on the sparsity of the underlying signals, to enable sampling below the classical Nyquist rate. The advantages with respect to the above “classical” technique is to transfer complexity from the acquisition phase to the reconstruction phase, moving the complexity where a higher amount of energy is available. The implementation of an ADC based on CS in a 0.18um CMOS technology by following a synergetic design between algorithm-circuit-system confirms this possibility.
Another example is given by lightweight methods for anomaly detection based on generalized spectral analysis. These are fundamental in any autonomous systems, such as unmanned vehicles or even space satellites. Methods based on generalized spectral analysis can be used for monitoring the signal energy laying along with the principal and anti-principal signal subspaces and call for an anomaly when such energy changes significantly with respect to normal conditions. Streaming approaches are for the online estimation of the needed subspaces are fundamental to avoid the complexity required by the standard full spectral analysis.