ENERGY EFFICIENT AND LOW POWER SIGNAL PROCESSING IN

WEARABLE MONITORING SYSTEMS FOR BIOMEDICAL

APPLICATIONS

ABSTRACT

Among the many commonalities shared between personal biomedical applications, the need to
process parallel streams of data in real-time is a dominating feature. This often requires a large
number of digital signal processing (DSP) and machine learning (ML) techniques. DSP is often
used to extract useful representations of the input data while machine learning is needed to
perform automated classification for diagnostic and detection purposes.

Wearable personal health monitoring systems can offer a cost effective solution for human
healthcare. These systems must provide both highly accurate, secured and quick processing and
delivery of vast amount of data. A tiny, energy efficient and domain-specific many core
accelerator referred to as power-efficient nano-clusters (PENC) is proposed to map and execute
the kernels of these applications. The PENC is able to reduce energy consumption by up to 80%
and 25% for DSP and ML kernels, respectively, when optimally parallelized.

The three compute-intensive personalized biomedical applications, namely, multichannel seizure

detection, multi physiological stress detection, and standalone tongue drive system (sTDS), to
evaluate the proposed many core performance relative to commodity embedded CPU, graphical
processing unit(GPU), and field programmable gate array (FPGA)-based implementations. The
application PENC improves throughput and energy efficient respectively.

M.SUNANDHA

18331D5710