BORACLE - Intelligent Algorithms

📌 Project Overview

Faculty Advisor: Dr. Nhut Ho & Dr. Xunfei Jiang
Institution: Autonomy Research Center for STEAHM (ARCS)
Dates of Research: August 2024 - Present
Funding & Support: CSUN Office of Undergraduate Research (OUR) - PODER, ARCS

View the official ARCS page on this research.

Abstract

Wearable devices and body sensor technology have evolved rapidly in the past years. While various types of devices collect different types of data, these data are currently only available to the device manufacturers’ internal team of developers. If the device’s manufacturers could make their devices’ body vital data available to 3rd party developers, significantly more health condition prognostics can be derived from trustworthy and intelligent algorithmic analysis. The decoupling of body sensor devices and intelligent algorithmic analysis of body vital data will accelerate independent and innovative growth in each of the areas. In this research, we will conduct comprehensive market research of commercial wearable devices, the APIs, the different collected data, and their digital health applications. We propose to build a platform, Boracle, that will collect data from any wearable devices, integrate the data and use big data technologies for intelligent analysis, and provide supporting services for both individual users and help app developers to create health condition prognostics. Borable will facilitate the intereoperability between applications and wearable body sensor devices and help end-users to improve their medical IQ/EQ and democratize accessibility to medical prognostics. [1]

🎯 Objectives

Develop machine learning models to analyze physiological signals from wearables
Lead a research team working on sub-problems such as stress prediction and activity modeling
Build and evaluate models for sleep stage classification using smartwatch data

🧩 Subtopics I Oversee

As the team lead, I mentor and support students working on the following sub-projects within the BORACLE Intelligent Algorithms initiative:

Stress Prediction from Sleep Patterns: Investigating correlations between sleep quality and daily stress levels using wearable data.
Injury Forecasting and Risk Prevention: Building models to detect signs of overtraining from activity metrics for soccer, competitive running, and football.
Physical Activity Analysis: Characterizing daily activity levels and patterns to support personalized health insights.
Smart Device Data Integration: Standardizing and preprocessing data from different consumer wearables for modeling.
Sleep Stage Classification (my own research focus): Developing algorithms to classify sleep stages (e.g., REM, light, deep) using time-series physiological signals.

🧠 Big Research Questions

Can we infer mental and physical health states using non-invasive wearable data alone?
How can we balance data accessibility, privacy, and performance for health forecasting?
What features most accurately distinguish different sleep stages using consumer-grade data?

🖼️ Posters

📦 Deliverables

Paper in-progress: “Stress Detection Through Sleep-Based Physiological Signals: An Integrated Semi-Supervised and Supervised Learning Approach”
Poster presentation for CSUNposium 2025
Poster presentation for 2024 CSUN REU Workshop

📈 Outcomes

Established baseline model for sleep stage classification from smartwatch metrics
Demonstrated the feasibility of decentralized team-led research under a unified goal
Mentored new graduate and undergraduate researchers in machine learning and research design

🔁 Ongoing / Future Work

Incorporating synthetic data to augment sleep datasets
Drafting first-author paper on sleep stage classification with wearable sensors

Brandon S. Ismalej