Energy-Efficient Hardware Architecture of Self-Organizing Map for ECG Clustering in 65-nm CMOS

摘要

An energy-efficient hardware architecture of a self-organizing map (SOM) for ECG clustering is proposed. It detects an R-peak, reconstructs the QRS complex around it, and clusters the complex by calculating the Euclidean distance between the complex and the weight vectors of each cell in the SOM network (i.e., 5×5 cells). In the operation mode, the cluster ID related to the minimum Euclidean distance is provided, while the tagged weight vectors are updated in the learning mode. The proposed SoC is 1735 × 1020 μm in CMOS 65-nm LP, and it consumes 5.853 mW at VDD = 1.2 V.