The increasing use of wearables in smart telehealth generates heterogeneousmedical big data. Cloud and fog services process these data for assistingclinical procedures. IoT based ehealthcare have greatly benefited fromefficient data processing. This paper proposed and evaluated use of lowresource machine learning on Fog devices kept close to the wearables for smarthealthcare. In state of the art telecare systems, the signal processing andmachine learning modules are deployed in the cloud for processing physiologicaldata. We developed a prototype of Fog-based unsupervised machine learning bigdata analysis for discovering patterns in physiological data. We employed IntelEdison and Raspberry Pi as Fog computer in proposed architecture. We performedvalidation studies on real-world pathological speech data from in homemonitoring of patients with Parkinson's disease (PD). Proposed architectureemployed machine learning for analysis of pathological speech data obtainedfrom smartwatches worn by the patients with PD. Results showed that proposedarchitecture is promising for low-resource clinical machine learning. It couldbe useful for other applications within wearable IoT for smart telehealthscenarios by translating machine learning approaches from the cloud backend toedge computing devices such as Fog.
展开▼
