Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper explores energy-efficient methods for arrhythmia detection in wearable devices by employing approximation techniques to reduce the power consumption of deep learning models. By integrating data precision reduction and approximate multiplication into a state-of-the-art deep learning architecture, the authors achieve a significant reduction in power and energy usage while maintaining high classification accuracy. Their proposed architecture demonstrates a 64.9% reduction in power consumption and a 61.5% reduction in energy consumption compared to existing solutions, making it viable for long-term monitoring in wearable applications.
Achieving a 64.9% reduction in power consumption for arrhythmia detection could revolutionize the feasibility of long-term monitoring in wearable health devices.
Cardiovascular diseases are the leading cause of death worldwide, and conditions such as arrhythmia often require long-term monitoring for effective detection and diagnosis. However, current wearable monitoring devices are bulky, uncomfortable, and typically rely on clinicians to manually evaluate electrocardiograms (ECGs). While Deep Learning (DL) algorithms have shown superior performance in arrhythmia detection and classification, their computational complexity coupled with high power consumption limit deployment in wearable devices. To address this challenge, this paper investigates the use of approximation techniques to reduce the power and energy consumption of DL architectures while maintaining acceptable classification performance. Specifically, techniques such as data precision reduction and approximate multiplication are investigated in a state-of-the-art DL model and its corresponding hardware architecture. The model is trained and validated using the MIT-BIH Arrhythmia Database, and hardware implementations employing various approximate multipliers are synthesized and evaluated. Compared with the state-of-the-art 8.75 {\mu}W (and 2.08 {\mu}J) reference architecture, our proposed architecture consumes 3.07 {\mu}W (and 2.17 {\mu}J) at 12 kHz, showing 64.9% reduction in power consumption while providing an acceptable output quality, i.e., 93.7% classification accuracy and 92.1% sensitivity. At 100 MHz, our proposed architecture consumes 9.45 mW (and 0.8 {\mu}J), showing 61.5% reduction in energy consumption as compared to the state-of-the-art architecture. These results demonstrate that our proposed approximations significantly extend wearable device battery life while preserving the required arrhythmia classification performance.