Nexus: A Low-Cost Wearable AI Health Companion Using IoT, Edge AI and a Smartphone-Hosted Voice Assistant

Abstract

Elderly people and patients with long-term conditions benefit from continuous, low-effort monitoring, yet most commercial health wearables are costly, screen-dependent and passive. This paper presents Nexus (Neural-Enabled eXtended User Support), a low-cost wearable AI health companion built on a three-layer IoT architecture. A wrist-worn ESP32-S3 device senses heart rate, blood oxygen saturation, motion and skin temperature and runs lightweight edge AI. A smartphone acts as the AI gateway and hosts speech recognition, a conversational assistant and speech synthesis. A caregiver interface provides status, alerts, history and location. The device has no display; it uses two-way voice, haptic feedback and a single SOS/TALK button. A proactive, health-focused assistant with short-term memory explains alerts in simple language, and a graded alert scheme with an emergency verification step reduces false alarms before caregivers are notified. The paper describes the requirements, hardware, software and AI design, privacy considerations, development methodology and a planned evaluation. Nexus is a health-monitoring and assistance prototype and is not a medical diagnostic device.

Introduction

1. Introduction

Healthcare is gradually moving towards continuous and home-based health monitoring using the Internet of Things (IoT). Wearable devices allow users to monitor important health parameters and physical activities throughout the day.

These devices are particularly useful for elderly people, individuals living alone, and patients with long-term health conditions. However, existing wearable devices are often expensive, difficult for older people to operate, and dependent on small displays. They also generate false alarms and provide limited interaction with users.

To overcome these limitations, the proposed project introduces Nexus (Neural-Enabled eXtended User Support), a low-cost AI-based wearable health monitoring and assistance system.

2. Problem Statement and Motivation

Elderly people and individuals with health conditions may experience emergencies such as falls, fainting, or sudden changes in health when caregivers are not present.

Existing monitoring devices have several limitations:

• High cost and complex operation.

• Dependence on small screens and touch interfaces.

• Limited voice interaction and health guidance.

• False emergency alerts due to simple threshold-based monitoring.

• Privacy concerns caused by transmitting raw health data to cloud servers.

The Nexus system aims to provide affordable, continuous health monitoring with voice assistance, local data processing, and reliable emergency notifications.

3. Literature Review

Previous research related to the proposed system can be divided into four main areas:

• Wearable Health Monitoring: Photoplethysmography (PPG) sensors are used to measure heart rate and blood oxygen saturation (SpO?). Accelerometers and gyroscopes help recognize physical activities and detect falls.

• IoT-Based Healthcare: IoT connects wearable sensors, smartphones, cloud platforms, and caregiver applications for continuous health monitoring.

• Edge AI and TinyML: Machine learning models are implemented on microcontrollers to detect unusual patterns locally, reducing internet dependency, power consumption, and data transmission.

• Conversational AI: Large Language Models (LLMs) enable natural voice communication and provide explanations and reminders. However, critical medical decisions should not depend entirely on AI-generated responses.

4. Research Gap

Existing healthcare devices generally focus on individual functions such as health monitoring, activity detection, voice assistance, or emergency alerts.

The proposed Nexus system addresses these limitations by combining:

1. Continuous monitoring of important health parameters.

2. On-device anomaly detection using Edge AI.

3. Display-free voice-based interaction.

4. Proactive AI assistance and health-related explanations.

5. Emergency verification before notifying caregivers.

6. A smartphone application for caregiver monitoring.

This combination aims to provide an affordable and user-friendly healthcare assistance solution.

5. Objectives of the Project

The main objectives of Nexus are:

1. To continuously monitor heart rate, blood oxygen level (SpO?), motion, and skin temperature trends.

2. To detect abnormal health patterns and possible emergencies using Edge AI.

3. To provide two-way voice communication between the user and the device.

4. To notify caregivers during emergencies and share the user's location.

5. To develop an affordable, lightweight, and comfortable wearable device.

6. To reduce false alarms through intelligent anomaly detection and emergency verification.

6. Scope of the Project

The first version (V1) of Nexus focuses on six major features:

• Heart rate and SpO? monitoring.

• Physical activity recognition.

• On-device anomaly detection.

• Voice-based interaction.

• Emergency verification and SOS alerts.

• Android-based caregiver application.

The system is designed as a health-monitoring and assistance prototype, not as a medical diagnostic device. ECG monitoring, onboard GPS, and cellular connectivity are planned for future versions.

7. System Requirements

Functional requirements:

• Measure heart rate and SpO? continuously.

• Monitor motion and skin temperature trends.

• Recognize activities and detect unusual patterns.

• Support voice communication through a wake word and button.

• Activate SOS through a long button press.

• Cancel an SOS through a double button press.

• Notify caregivers with the user's location if emergency verification fails.

Non-functional requirements:

• Low-cost and easily available components.

• Lightweight and comfortable wrist-worn design.

• Long battery life through efficient power management.

• Local data storage during temporary network failures.

• Privacy protection through local data processing.

8. Proposed System Architecture

The Nexus system consists of three main layers:

• Wearable Layer: Collects health and activity information through sensors and performs basic data processing and anomaly detection using Edge AI.

• Smartphone Layer: Acts as a communication gateway and provides advanced AI services, voice interaction, and internet connectivity.

• Caregiver Application Layer: Displays health information, emergency notifications, and user location to family members and caregivers.

The system uses local processing for important health-event detection, while the smartphone handles more complex AI operations.

9. Working Principle

The wearable device continuously collects physiological and activity data using sensors. The microcontroller processes the readings and identifies unusual patterns.

When a possible emergency is detected, the device communicates with the user through voice or vibration and asks for confirmation.

• If the user confirms that they are safe, no emergency notification is sent.

• If the user does not respond or requires help, an emergency alert is sent to registered caregivers through the smartphone application.

• The caregiver receives relevant information and the user's location.

The AI assistant also provides voice-based explanations, reminders, and general health-related assistance.

Conclusion

This paper presented Nexus, a low-cost wearable AI health companion based on a three-layer IoT architecture. By combining a display-free wrist device, on-device anomaly filtering, a smartphone-hosted proactive voice assistant and a caregiver app, Nexus addresses the cost, accessibility and passivity problems of many existing wearables. The design keeps safety-critical decisions in deterministic and edge-AI logic, uses a graded alert scheme and adds a confirmation step before alerting caregivers. The next stage of the work is to complete the prototype, carry out the planned evaluation and report measured results.

References

[1] J. Gubbi, R. Buyya, S. Marusic, and M. Palaniswami, “Internet of Things (IoT): A vision, architectural elements, and future directions,” Future Gener. Comput. Syst., vol. 29, no. 7, pp. 1645–1660, Sep. 2013. [2] S. C. Mukhopadhyay, “Wearable sensors for human activity monitoring: A review,” IEEE Sensors J., vol. 15, no. 3, pp. 1321–1330, Mar. 2015. [3] J. Allen, “Photoplethysmography and its application in clinical physiological measurement,” Physiol. Meas., vol. 28, no. 3, pp. R1–R39, Mar. 2007. [4] Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology, “Heart rate variability: Standards of measurement, physiological interpretation, and clinical use,” Circulation, vol. 93, no. 5, pp. 1043–1065, Mar. 1996. [5] P. Warden and D. Situnayake, TinyML: Machine Learning with TensorFlow Lite on Arduino and Ultra-Low-Power Microcontrollers. Sebastopol, CA, USA: O’Reilly Media, 2019. [6] R. David et al., “TensorFlow Lite Micro: Embedded machine learning for TinyML systems,” in Proc. Mach. Learn. Syst. (MLSys), vol. 3, 2021. [7] S. Hymel et al., “Edge Impulse: An MLOps platform for tiny machine learning,” in Proc. Mach. Learn. Syst. (MLSys), 2023. [8] T. B. Brown et al., “Language models are few-shot learners,” in Adv. Neural Inf. Process. Syst. (NeurIPS), vol. 33, 2020, pp. 1877–1901. [9] MAX30102: High-Sensitivity Pulse Oximeter and Heart-Rate Sensor for Wearable Health, datasheet, Maxim Integrated. [10] BMI270: Small, Low-Power Inertial Measurement Unit, datasheet, Bosch Sensortec. [11] MAX30205: Human Body Temperature Sensor, datasheet, Maxim Integrated. [12] ESP32-S3 Series Datasheet, Espressif Systems. [13] MAX98357A/MAX98357B: Tiny, Low-Cost, PCM Class D Amplifier, datasheet, Maxim Integrated. [14] Bluetooth Core Specification, Bluetooth Special Interest Group. [15] Getting Started with Seeed Studio XIAO ESP32S3 Sense, Seeed Studio documentation. [16] Bluetooth Low Energy Overview, Android Developers documentation, Google. [17] Firebase Documentation, Google.

Copyright

Copyright © 2026 Suyog Danole, Ayan Mulla, Sarvesh Teli, Arif Shirgave, Sairaj Salunkhe. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.