A look at signal analysis in resource-constrained medical-sensor applications

Given the increasing emphasis on data-driven medicine, this overview paper explores how healthcare functions can be extended to on-patient sensor platforms. While the functions that will be of interest in this context are not precisely known, the approaches within data-driven medicine are becoming clear. This guides us to the algorithms that will be used for diagnosis, monitoring, and therapy. In particular, there is an emphasis on machine-learning algorithms, which enable the creation of robust models for signal analysis from data. Focusing on such algorithms, this paper looks first at digital platforms for signal analysis that provide adequate programmability while addressing computational energy. Then, it considers approaches to signal acquisition that focus on acquiring the specific information needed within algorithms for signal analysis, with the aim that such focus relaxes both the specifications for mixed-signal interfaces and the subsequent computations required. The ideas presented reference hardware prototype demonstrations.