Intelligent Edge Biomedical Sensors in the Internet of Things (IoT) Era

Elisabetta De Giovanni, Farnaz Forooghifar, Gregoire Surrel, Tomás Teijeiro, Miguel Peon, Amir Aminifar, David Atienza AlonsoComputer 2022 · DOI · OpenAlex

Abstract

The design of reliable wearable systems for real-time and long-term monitoring presents major challenges, although they are poised as the next frontier of innovation in the context of the Internet-of-Things (IoT) to provide personalized healthcare. This new generation of biomedical sensors targets to be interconnected in ways that improve our lives and transform the medical industry. However, several key challenges remain in achieving this potential due to the inherent resource-constrained nature of wearable systems for Big Data medical applications. This paper discusses the opportunities for edge computing and edge AI in next-generation intelligent biomedical sensors, introducing the notion of self-awareness to trade-off machine-learning performance versus system lifetime.

Introduction: The Energy-Accuracy Paradox

Wearable biomedical sensors face a fundamental "energy-accuracy" paradox. To detect life-threatening pathologies—such as epileptic seizures or cardiac arrhythmias—in real-time, devices require complex machine learning models and high-frequency data sampling. However, these devices are severely resource-constrained, relying on small batteries that must last for extended periods.

Traditional designs are typically "worst-case" oriented: they sample at high rates and run complex models constantly, even when the patient is healthy and the signal is redundant. To move from simple gadgets to reliable clinical tools, wearables must move beyond static configurations and embrace self-awareness—the ability of a system to monitor its own state and environment to make real-time trade-offs between energy consumption and diagnostic accuracy.

The Self-Aware Framework

The proposed framework for intelligent edge sensors adapts across three distinct layers: application, architecture, and hardware.

1. Application Layer: Two-Level Learning

Instead of running a single, computationally expensive model, the system employs a hierarchical approach. A "Simple Model" handles routine monitoring. A secondary "Confidence Model" evaluates the output; if the simple model is unsure, the system triggers a "Complex Model" (either locally or offloaded to the cloud/fog).

Figure 6: Motivational example of the self-awareness concept. The shaded region represents the uncertainty zone where a simple classifier lacks confidence, triggering the need for a more complex analysis or higher resource allocation.

By identifying when a simple model is confident, the system can stay in a low-power state for the majority of the time, as most medical data during routine monitoring is "normal."

2. Architecture Layer: Platform-Awareness

At the architectural level, the research explores ultra-low-power (ULP) multi-core platforms, such as the RISC-V based GAP8. By parallelizing tasks across multiple cores at lower clock frequencies and voltages, the system achieves higher energy efficiency than single-core processors. This "parallelism for energy" approach allows the system to finish tasks faster and enter deep sleep states sooner, or run at a lower voltage where transistors operate more efficiently.

3. Hardware Layer: Event-Driven Sampling

Moving away from uniform Shannon-Nyquist sampling, the authors implement signal-driven sampling. Using techniques like Polygonal Approximation (e.g., the Wall-Danielsson algorithm), the sensor only records data points when the signal deviates from a linear path. This effectively "ignores" the steady, redundant parts of a heartbeat, drastically reducing the data load at the source.

The Self-Awareness Loop

Central to this architecture is a closed-loop system that allows the device to adapt to the real world much like a living organism. The system observes its environment and internal state, analyzes the requirements, and acts to optimize its operation.

Figure 3: The self-awareness loop (Observe-Analyze-Act) integrated into the system architecture to manage the trade-off between performance and energy efficiency.

Key Results and Impact

The researchers validated these strategies using the MIT-BIH Arrhythmia Database and real-life epileptic seizure detection datasets. Key findings include:

  • Seizure Detection Efficiency: The self-aware classifier achieved 80.83% performance (nearly matching the 82.53% of the complex model) while consuming only 39.4% of the energy.
  • Sampling Reduction: For ECG monitoring, the self-aware sampling strategy reduced the effective sampling rate from 360 Hz to 13.6 Hz (a ~26x reduction) while maintaining a nearly perfect F1-score (99.73%) for heartbeat detection.
  • Offloading Benefits: Offloading complex tasks to a "Fog" device (like a smartphone) reduced the wearable's energy consumption for classification from 13.11 mJ to 3.65 mJ.

Conclusion

The transition to intelligent edge biomedical sensors requires a paradigm shift from design-driven to signal-driven systems. By integrating self-awareness across the application, architecture, and circuit levels, next-generation wearables can provide clinical-grade monitoring while operating within the strict energy envelopes of the IoT era. Future work will focus on generalizing these adaptive thresholds across more diverse and noisy bio-signals, such as PPG or EMG, during intense physical activity.

Summary updated

Abstract

Abstract The design of reliable wearable systems for real-time and long-term monitoring presents major challenges, although they are poised as the next frontier of innovation in the context of Internet-of-Things (IoT) to provide personalized healthcare. This new generation of biomedical sensors targets to be interconnected in ways that improve our lives and transform the medical industry. Therefore, they offer an excellent opportunity to integrate the next generation of artificial intelligence (AI) based techniques in medical devices. However, several key challenges remain in achieving this potential due to the inherent resource-constrained nature of wearable systems for Big Data medical applications, which need to detect pathologies in real time. Concretely, in this chapter, we discuss the opportunities for edge computing and edge AI in next-generation intelligent biomedical sensors in the IoT era and the key challenges in wearable systems design for pathology detection and health/activity monitoring in the context of IoT technologies. First, we introduce the notion of self-awareness toward the conception of the next-generation intelligent edge biomedical sensors to trade-off machine-learning performance versus system lifetime, according to the application requirements of the medical monitoring systems. Subsequently, we present the implications of personalization and multi-parametric sensing in the context of the system-level architecture of intelligent edge biomedical sensors. Thus, they can adapt to the real world, as living organisms do, to operate efficiently according to the target application requirements and available energy at any moment in time. Then, we discuss the impacts of self-awareness and low-power requirements at the circuit level for sampling through a paradigm shift to react to the input signal itself. Finally, we conclude by highlighting that the techniques discussed in this chapter may be applied jointly to design the next-generation intelligent biomedical sensors and systems in the IoT era.
BibTeX
@inproceedings{giovanni2022intelligent,
  author = {Elisabetta De Giovanni and Farnaz Forooghifar and Gregoire Surrel and Tomás Teijeiro and Miguel Peon and Amir Aminifar and David Atienza Alonso},
  title = {{Intelligent Edge Biomedical Sensors in the Internet of Things (IoT) Era}},
  booktitle = {Computer},
  year = {2022},
  doi = {10.1007/978-981-16-7487-7\_13},
}
Figure