One Signal Reads Blood Pressure and Recognizes Gestures: The X-Sig Skin Patch Layers “Electricity” and “Force” Together

Wearable devices that seek to understand the heart, blood vessels, and muscles simultaneously usually require multiple sensors to collect data separately before the circuitry integrates them. A research team from the National University of Singapore and other institutions has proposed another approach: an X-Sig sensor attached to the skin first combines biopotential and mechanical signals into a single waveform, after which algorithms extract the physiological features within it.

The design combines a conductive, adhesive, and recyclable dry electrode with a 28-micrometre-thick perforated piezoelectric film. The electrode receives electrocardiogram or electromyogram signals, while the piezoelectric layer senses pressure changes caused by the radial artery pulse or muscle contractions. The two types of signals are fused at the sensor, requiring only one output channel and one analogue front end. However, the study demonstrates two configurations—electrocardiogram plus pulse, and electromyogram plus muscle-force signals—rather than simultaneously resolving all four signals in every scenario.

In the cardiovascular demonstration, the system identified electrocardiographic R waves and pulse peaks from the composite waveform, calculated heart rate and pulse arrival time, and then used supervised machine learning to estimate systolic and diastolic blood pressure. Results reported in the preprint show that, compared with a commercial electronic blood pressure monitor, the difference was −0.43±6.11 mmHg for systolic blood pressure and 0.96±3.51 mmHg for diastolic blood pressure. Healthy volunteers also wore the device during sit-to-stand transitions, the Valsalva manoeuvre, and brief exercise to test continuous changes in haemodynamic parameters.

The researchers also placed the patch on the forearm and used a convolutional neural network to distinguish ten hand movements. The methods data show that 70 signal sets were obtained for each movement, with 80% used for training and 20% for testing. After electromyographic and muscle-force signals were fused, classification accuracy reached 96.4%, compared with 72.1% using electromyography alone and 82.9% using muscle-force signals alone. This comparison supports the idea that information from different physical sources can complement one another, but the dataset is small and remains insufficient to demonstrate that the model can maintain the same performance across users and wearing positions.

The practical value of single-channel fusion extends beyond reducing the number of wires. If it can reduce the number of analogue front ends, the volume of transmitted data, and subsequent computation, the patch could potentially be made smaller and more energy-efficient, with applications ranging from tracking blood pressure trends at home to recognizing rehabilitation gestures or supporting human–machine interfaces. The paper has been peer-reviewed and published in *Nature Sensors*, and the source data, supplementary materials, and code have been made public, allowing other teams to examine its analytical workflow.

However, the current research does not support treating X-Sig as a replacement for a clinical blood pressure monitor. The published results do not provide validation in a large participant population sufficiently representative of different ages, skin conditions, cardiovascular diseases, and long-term daily activities. Nor do they answer questions about individual calibration, perspiration and displacement, tolerance of prolonged adhesion, or reproducibility across devices. In particular, if cuffless blood pressure estimation is to be used for medical decision-making, it must still undergo broader, independent, and prospective clinical evaluation in accordance with formal standards.

The technology is also covered by an intellectual property strategy: four authors disclosed that they are inventors on a pending Chinese patent. At present, X-Sig is better understood as a working prototype that advances the engineering concept of “fuse first, decode later” for multimodal wearable sensing. Whether it can become a practical medical device will depend not only on laboratory accuracy, but also on whether its performance can be reproduced reliably in real-world populations and during long-term use.