TLDR: PulseFi is a new, low-cost, non-intrusive system that uses standard Wi-Fi signals (Channel State Information) and a custom Long Short-Term Memory (LSTM) neural network to accurately monitor heart rate, breathing rate, and detect apnea events. It operates on commodity hardware like ESP32s and Raspberry Pis, achieving high clinical accuracy and robust performance across varying distances and participant positions, making vital sign monitoring more accessible and affordable for home and elderly care.
Monitoring vital signs like heart rate, breathing rate, and detecting conditions such as sleep apnea is crucial for healthcare, especially with an aging global population and rising chronic diseases. Traditional methods often involve contact-based sensors like pulse oximeters or chest straps, which can be uncomfortable and impractical for long-term use. Camera-based systems, while non-contact, raise privacy concerns and struggle in low-light conditions or when subjects are obstructed.
Introducing PulseFi: A New Era of Non-Intrusive Monitoring
A novel system called PulseFi is emerging as a promising solution, offering a low-cost, non-intrusive, and highly accurate way to monitor cardiopulmonary activity and detect apnea events. Developed by Pranay Kocheta, Nayan Sanjay Bhatia, and Katia Obraczka, PulseFi leverages ubiquitous Wi-Fi signals and artificial intelligence to provide continuous health insights. You can delve deeper into their work by reading the full research paper: PulseFi: A Low Cost Robust Machine Learning System for Accurate Cardiopulmonary and Apnea Monitoring Using Channel State Information.
How PulseFi Works Its Magic
PulseFi operates using readily available, inexpensive Wi-Fi devices, such as ESP32s or Raspberry Pi 4s. One device acts as a transmitter, continuously sending Wi-Fi packets, while another acts as a receiver. When a person is positioned between these devices, their subtle chest movements from breathing and heartbeats cause tiny modulations in the Wi-Fi signal. This information is captured as Channel State Information (CSI), which essentially describes how the Wi-Fi signal propagates through the environment.
The system then employs a sophisticated signal processing pipeline. First, it extracts the amplitude (strength) of the CSI data, deliberately avoiding complex phase information, which allows it to work effectively with single-antenna, low-cost hardware. Next, it removes static noise and filters the signal to isolate the specific frequencies associated with heartbeats (0.8 to 2.17 Hz) and breathing (0.1 to 0.5 Hz). A smoothing filter further refines the data, which is then segmented and normalized before being fed into the AI component.
The core of PulseFi’s intelligence lies in its custom, low-compute Long Short-Term Memory (LSTM) neural network. LSTMs are a type of deep learning model particularly adept at processing sequential and temporal data, making them ideal for analyzing the continuous stream of Wi-Fi signal variations. This AI model learns to identify patterns corresponding to heart rate, breathing rate, and the absence of breathing during apnea events, even in noisy environments.
Unpacking the Advantages
PulseFi addresses several critical requirements for vital sign monitoring systems:
- Accessibility and Cost-Effectiveness: By using commodity Wi-Fi devices, PulseFi significantly lowers the barrier to entry, making advanced monitoring more widely available.
- Robustness: The system maintains consistent accuracy across varying distances (1m to 3m) and different participant positions, a significant improvement over many existing Wi-Fi sensing methods that often degrade with distance or movement.
- Reliability: Evaluated using two diverse datasets, including one with 118 participants, PulseFi demonstrates strong generalization capabilities.
- Computational Efficiency: Its compact LSTM models are designed to run on resource-constrained devices like the ESP32 and Raspberry Pi, enabling real-time, continuous monitoring without requiring powerful computing hardware.
- Accuracy: PulseFi delivers clinically relevant accuracy for heart rate, breathing rate, and apnea detection, often outperforming more specialized and expensive multi-antenna systems.
Also Read:
- HiMAE: A Compact AI Framework for Smartwatches That Deciphers Wearable Health Data
- EcoScale-Net: A Breakthrough in Lightweight ECG Classification for Cardiac Health
Impressive Performance Metrics
The research highlights PulseFi’s remarkable accuracy. For heart rate monitoring, it achieved a Mean Absolute Error (MAE) as low as 0.17 BPM on the EHealth dataset, with over 99% of estimates falling within the clinical threshold of 1.5 BPM error. For breathing rate, it achieved an MAE of 0.09 breaths/min, with over 99% of estimates within 0.75 breaths/min error. Apnea detection was nearly perfect, boasting 99.4% accuracy, 95.7% sensitivity, and 99.7% specificity, along with a high Kappa coefficient of 95.9, indicating excellent agreement with actual events.
These results underscore PulseFi’s potential to transform vital sign monitoring, moving it from specialized clinical settings to everyday environments like homes and long-term care facilities. Its ability to provide accurate, non-intrusive, and affordable health insights could significantly improve early detection of health issues and enhance continuous care for millions worldwide.
