Machine-learning specialist Edge Impulse has introduced capabilities for heart rate (HR) and heart rate variability (HRV) data processing that is said to outperform existing algorithms and enable higher accuracy interpreting PPG and ECG sensor outputs.

HR and HRV are important measurements for numerous health and medical endeavours. With the HR and HRV blocks, enterprises and developers can better tackle problems involving activity and sleep tracking, fall detection, sleep monitoring, and atrial fibrillation.

“Our new algorithm generates clean HR and HRV values from a PPG sensor via our augmentation of standard processing techniques,” said Alex Elium, lead DSP engineer at Edge Impulse. “Our enhancements mitigate significant noise typically associated with using data from wearables, such as on a finger, outperforming known algorithms in MAE, variance and resource usage. This significantly reduces the R&D investment to build custom algorithms that would take years to refine for use in the field, whether for clinical trials or consumer wearable devices.”

The launch is part of Edge Impulse’s broader suite of edge AI tools for health-related use cases that can be deployed to any target with algorithms optimised for any hardware, from high power GPUs down to the smallest MCUs on wearables. On-device intelligence can transform health devices with real-time insights, enhanced privacy and extended battery life.

Numerous health data types now have pre-built algorithms in Edge Impulse, including a set of HRV algorithms optimised for on-device performance to track accelerating activity, falls, sleep, stress and atrial fibrillation. The toolkit is said to use the smallest footprint on the market with 16x less RAM, with the smallest mean error compared with current industry standards.

Other health data types using this include: ECG and PPG for monitoring heart health in real time; EEG to capture brainwave patterns for neurological assessments; motion to track movement and posture; and body temperature monitoring for health changes.

A research data lake is included that can be useful to store, aggregate and validate all clinical trial data in one place, with the infrastructure to scale clinical studies to thousands of subjects. There is also a tool that provides real-time project monitoring capabilities and increases team efficiency for an accelerated project success.

Edge Impulse’s platform is designed to operate in compliance with recent FDA regulations around AI use for medical devices, making it a go-to choice for numerous innovators in the healthcare and wellness space.

Health companies using Edge Impulse’s medical-grade tools for edge AI include:

• Know Labs: An emerging developer of non-invasive medical diagnostics technology bringing the first FDA-approved, non-invasive blood glucose monitoring to market.
• Oura: The maker of popular health-tracking device Oura Ring uses Edge Impulse’s platform to build and train data models more quickly and at a higher accuracy for medical research and development.
• Nowatch: The wearables company works with Edge Impulse to use AI to monitor mental well-being. The company’s device uses a combination of sensors to measure cortisol, which then can be used to measure stress levels in human beings.
• Hyfe: Hyfe leverages Edge Impulse’s platform to deploy its AI model for cough detection to the edge, facilitating real-time and efficient monitoring of respiratory health.
• SlateSafety: Originally designed for firefighters, the company’s Band v2 wearable tracks signs of stress for people working in dangerous or otherwise extreme conditions. Edge Impulse helps SlateSafety aggregate and process sensor data on-device, protecting workers wherever they’re needed regardless of connectivity or signal strength.

Edge Impulse (edgeimpulse.com) offers machine-learning tooling, enabling enterprises to build smarter edge products. Its technology empowers developers to bring more AI products to market faster, and helps enterprises rapidly develop industry specifications in weeks instead of years. The California-based company provides automations and low-code capabilities to make it easier to build datasets and develop AI for edge devices.