New sensor from Singapore detects signs of stress

The main obstacle to using existing wearable devices, such as smartwatches, for clinical purposes is “motion artifacts.” This term refers to interference in the signal caused by muscle contractions, friction between the device and the skin, and sensor displacement. These noises interfere with the weak electrical signals from the heart, making it difficult to accurately measure heart rate variability, a key indicator of the autonomic nervous system’s functioning.

The developed MAP (Metahydrogel Artifact-mitigating Platform) solves the problem of noise not only by software methods, but also at the level of the material structure itself. The sensor is a soft, skin-like hydrogel that integrates self-organizing nanoparticles. These particles form periodic structures that can mechanically absorb and dissipate body vibrations, acting as a filter for physical interference in certain frequency ranges.

The second level of noise protection is provided by a biocompatible electrolyte based on glycerin and water. It controls the speed of ion movement within the gel in such a way that low-frequency heart signals (below 30 Hz) pass through unhindered, while high-frequency electrical noise from muscle activity is effectively suppressed. This combination of physical and chemical filtering allows for the purification of the signal before it enters the digital processing stage.

To further refine the data, the system employs machine learning algorithms that remove any remaining unstructured noise while preserving the essential characteristics of the electrocardiogram (ECG). As a result, the device demonstrates a signal-to-noise ratio (SNR) of 37.36 dB, which is significantly higher than that of commercial trackers, which typically have an SNR of 10-20 dB and drop sharply during movement.

The accuracy of blood pressure measurement using this sensor meets the international clinical standard ISO 81060-2, with a deviation of no more than 3 mmHg. Thanks to the high quality of the data obtained, the neural network is able to classify human fatigue levels with an accuracy of up to 92%. For comparison, when trained on data from conventional sensors, the accuracy of fatigue recognition does not exceed 64% due to the abundance of interference.

The MAP platform has mechanical properties similar to those of human biological tissues. It is elastic, durable under repeated stretching, and highly breathable, allowing the skin to “breathe” under the sensor. This makes the device suitable for long-term continuous wear for several days, which is crucial for monitoring the condition of drivers, pilots, and operators of complex systems.

The researchers successfully tested the system in a simulated driving environment, where they needed to identify moments when the body was in a state of critical fatigue. In addition to ECG and blood pressure, the platform demonstrated its effectiveness in filtering out noise during the recording of heart tones, breathing, voice, and even brain waves (EEG). This highlights the potential of the technology for creating comprehensive mental health monitoring systems that can detect early signs of depression and anxiety disorders.

Currently, the NUS team is working on standardizing the manufacturing process to transition the production of sensors from laboratory settings to an industrial scale. In the near future, we plan to work closely with medical practitioners to establish clearer correlations between biological signals and specific pathological conditions. This will enable us to transform the wearable sensor into a fully-fledged diagnostic tool that can be used in clinical practice.