Sorry, you need to enable JavaScript to visit this website.

Low-Complexity Compressed Analysis in Eigenspace with Limited Labeled Data for Real-Time Electrocardiography Telemonitoring

Primary tabs

Citation Author(s):
Kai-Chieh Hsu, Bo-Hong Cho, Ching-Yao Chou, and An-Yeu (Andy) Wu
Submitted by:
Kai-Chieh Hsu
Last updated:
12 December 2018 - 10:09pm
Document Type:
Presentation Slides
Document Year:
Presenters Name:
Kai-Chieh Hsu
Paper Code:



To achieve real-time electrocardiography (ECG) telemonitoring, we need to overcome the scarce bandwidth. Compressed sensing (CS) emerges as a promising technique to greatly compress ECG signal with little computation. Furthermore, with edge-classification, we can reduce the data rate by transmitting abnormal ECG signals only. However, there are three main limitations: limited number of labeled ECG signal, tight battery constraint of edge devices and low response time requirement. Task-driven dictionary learning (TDDL) appears as an appropriate classifier to render low complexity and high generalization. Combining CS with TDDL directly (CA-N) will degrade classification and need higher complexity model. In this paper, we proposed an eigenspace-aided compressed analysis (CA-E) integrating principal component analysis (PCA), CS and TDDL, sustaining not only light complexity but high performance under exiguous labeled ECG data set. Simulation results show CA-E reduces about 67% parameters, 76% training time, 87% inference time and has the smaller accuracy variance to CA-N counterpart.

0 users have voted:

Dataset Files