Edge AI

Embedded Machine Intelligence Lab
Last updated on Feb 20, 2024

The stringent constrained resources available on tiny sensor nodes introduce a number of challenges regarding accuracy, power-efficiency, user-comfort, and security. These design considerations, however, often impose conflicting requirements. Thus, a comprehensive research approach to design future medical embedded systems and corresponding optimizations at different levels must consider these interdependent and conflicting requirements. We research methods of optimizing medical embedded systems for power-efficiency while taking into account other performance metrics. The goal is to develop tools, methodologies, and algorithms towards comprehensive approaches to address cross-layer optimization issues related to power, performance, user-comfort, and security.

Embedded Machine Intelligence Lab
Embedded Machine Intelligence Lab
Research Lab

At the Embedded Machine Intelligence Lab (EMIL) at Arizona State University, we develop next-generation AI, sensing, and digital health technologies that bridge the gap between algorithmic innovation and real-world clinical impact. Our research focuses on the design, development, and validation of robust, interactive, efficient, and trustworthy machine learning methods for real-world pervasive systems operating under dynamic and resource-constrained conditions.

Publications

Hybrid Attention Model Using Feature Decomposition and Knowledge Distillation for Blood Glucose Forecasting
A hybrid attention framework designed for accurate and efficient blood glucose forecasting using multimodal data using feature decomposition and knowledge distillation
IEEE Transactions on Mobile Computing (IEEE TMC) - April 2026
PDF Cite Code Project Project Project DOI
Hybrid Attention Model Using Feature Decomposition and Knowledge Distillation for Blood Glucose Forecasting
Minimum-Cost Channel Selection in Wearables
We introduce a framework for the channel selection problem, mathematically formulate the minimum-cost channel selection (MCCS), and propose two novel algorithms to solve the MCCS problem.
IEEE-EMBS International Conference on Body Sensor Networks: NextGen Health: Sensor Innovation, AI, and Social Responsibility (BSN'24)
PDF Project
Minimum-Cost Channel Selection in Wearables
Efficient Sensing and Classification for Extended Battery Life
We propose a resource-efficient, real-time human activity recognition framework, transforming the multi-class classification problem into a hierarchical model based on the Metabolic Equivalent of Task (MET), creating a personalized structure for each individual.
In Activity Recognition and Prediction for Smart IoT Environments, Springer - January 2024
Project
Efficient Sensing and Classification for Extended Battery Life
On-Device Machine Learning for Diagnosis of Parkinson's Disease from Hand Drawn Artifacts
CNN model developed, quantized, and deployed on Android platform to screen PD patients from hand-drawn spirals.
IEEE International Conference on Wearable and Implantable Body Sensor Networks (BSN’22)
PDF Project
On-Device Machine Learning for Diagnosis of Parkinson's Disease from Hand Drawn Artifacts
Probabilistic Cascading Classifier for Energy-Efficient Activity Monitoring in Wearables
A cascading approach for optimizing energy and memory requirements of a machine learning model running on an embedded device.
IEEE Sensors Journal, 2022.
PDF Project
Probabilistic Cascading Classifier for Energy-Efficient Activity Monitoring in Wearables
Optimal Policy for Deployment of Machine Learning Models on Energy-Bounded Systems
International Joint Conference on Artificial Intelligence (IJCAI), 2020
PDF Cite Project
Optimal Policy for Deployment of Machine Learning Models on Energy-Bounded Systems
Improved Knowledge Distillation via Teacher Assistant
AAAI Conference on Artificial Intelligence (AAAI), 2020
PDF Cite Code Project Video
Improved Knowledge Distillation via Teacher Assistant
Resource-Efficient Computing in Wearable Systems
IEEE Workshop on Smart Service Systems (SmartSys) in Conjunction with IEEE SmartComp 2019, June, 12, 2019, Washington D.C., USA
PDF Cite Project
Resource-Efficient Wearable Computing for Real-Time Reconfigurable Machine Learning: A Cascading Binary Classification
The 16th IEEE International Conference on Wearable and Implantable Body Sensor Networks (BSN 19), Chicago, IL, USA
PDF Cite Project
Toward Ultra-Low-Power Remote Health Monitoring: An Optimal and Adaptive Compressed Sensing Framework for Activity Recognition
IEEE Transactions on Mobile Computing (TMC)
PDF Cite Project
Toward Ultra-Low-Power Remote Health Monitoring: An Optimal and Adaptive Compressed Sensing Framework for Activity Recognition
Trading Off Power Consumption and Prediction Performance in Wearable Motion Sensors: An Optimal and Real-Time Approach
ACM Transactions on Design Automation of Electronic Systems (TODAES), vol. 23, no. 5, article 67, October 2018
Cite Project
Adaptive compressed sensing at the fingertip of Internet-of-Things sensors: An ultra-low power activity recognition
IEEE/ACM Design, Automation and Test in Europe (DATE), March 27-31, 2017, Lausanne, Switzerland
Cite Project
Adaptive compressed sensing at the fingertip of Internet-of-Things sensors: An ultra-low power activity recognition
A Hardware-Assisted Energy-Efficient Processing Model for Activity Recognition Using Wearables
ACM Transactions on Design Automation of Electronic Systems (TODAES), vol. 21, no. 4, pp.1-27, 2016
Cite Project
CyHOP: A generic framework for real-time power-performance optimization in networked wearable motion sensors
International Conference on Computer Design (ICCD), 2016
PDF Cite Project
CyHOP: A generic framework for real-time power-performance optimization in networked wearable motion sensors
Power-Aware Computing in Wearable Sensor Networks: An Optimal Feature Selection
IEEE Transactions on Mobile Computing (TMC), 14(4), pp.800-812, 2016
Cite Project