Periodic Reporting for period 1 - EU-TRAINS (Towards an Ecosystem of User-centric devices and services for multisport Training and Remote healthcare enabled by an Artificial Intelligence-based Network of Sensors)
Période du rapport: 2024-06-01 au 2025-11-30
In the contemporary landscape of healthcare and fitness, the integration of advanced technological solutions is imperative to address the challenges posed by an ageing society and the increasing demand for personalized, adaptable training programs. Consequently, a wide range of smart wearables and applications are available in the market today, including for example smart watches, smart rings, or swim trackers.
The main goal of the EU-TRAINS project is to leverage the smart wearables ecosystem by developing and demonstrating key innovations that target central challenges of current state-of-the-art technology, resulting in the following overall objectives:
1) Provide additional and accurate live insights to the users by applying real-time data fusion of multiple parameters using edge-AI techniques.
2) Extend real-time parameter analysis and user feedback to swimming applications by enabling reliable in-water communication of data between sensors, the central edge-AI processor, and the cloud environment.
3) Add biochemical parameters to the set of input data by realizing a wearable biosensor for non-invasive biochemical analysis of sweat.
4) Improve the usability and comfort of smart wearables by integrating the sensing & data processing hardware into durable and recyclable smart textiles.
5) Make sure that the developed technology is well accepted an adapted by end users by embodying the human factor in the design loop.
Furthermore, with its well-balanced pan-European consortium of industrial partners, SMEs, RTOs, & Universities, EU-TRAINS strengthens the position of Europe in wearable technology by creating a resilient made-in-Europe supply chain.
The main goal of the EU-TRAINS project is to leverage the smart wearables ecosystem by developing and demonstrating key innovations that target central challenges of current state-of-the-art technology, resulting in the following overall objectives:
1) Provide additional and accurate live insights to the users by applying real-time data fusion of multiple parameters using edge-AI techniques.
2) Extend real-time parameter analysis and user feedback to swimming applications by enabling reliable in-water communication of data between sensors, the central edge-AI processor, and the cloud environment.
3) Add biochemical parameters to the set of input data by realizing a wearable biosensor for non-invasive biochemical analysis of sweat.
4) Improve the usability and comfort of smart wearables by integrating the sensing & data processing hardware into durable and recyclable smart textiles.
5) Make sure that the developed technology is well accepted an adapted by end users by embodying the human factor in the design loop.
Furthermore, with its well-balanced pan-European consortium of industrial partners, SMEs, RTOs, & Universities, EU-TRAINS strengthens the position of Europe in wearable technology by creating a resilient made-in-Europe supply chain.
During the first project period, we carried out extensive stakeholder engagements to define the end-user requirements and most relevant use scenarios. To that end, we conducted five distinct online surveys targeting athletes, rehabilitation users, and swimming coaches. Furthermore, we performed expert interviews with rehabilitation specialists, summing up to almost 1300 evaluated stakeholder responses. Based on these results, we first defined the specific goals for our targeted edge-AI algorithms, which we concluded to be the following:
1) Land-based scenario: An edge-AI algorithm for predicting the personal performance indicator and metabolic thresholds in real-time during training by analysing the data streams of a multitude of wearable sensors, including also biochemical data delivered by our wearable sweat biosensor. The performance indicator and metabolic thresholds delineate exercise intensity domains and provide quantitative benchmarks for training prescription, and are, therefore, equally relevant both for athletic and rehabilitation use scenarios. However, a valid assessment of these parameters can currently only be obtained in a laboratory environment by invasive methods.
2) Water-based scenario: An edge-AI algorithm that provides live assessment of swimming techniques by real-time analysis of biomechanical parameters delivered by multiple sensors placed at various critical body positions. Here, currently available systems can only record data for post-training analysis, which is a direct result of the non-existence of effective in-water data communication technology. By leveraging the in-water communication technology developed in the project, we target to deliver for the first time live swimming technique assessment, thereby enabling direct response to swimmers and swimming trainers.
With the edge-AI algorithms defined, we next specified distinct demonstrator Setups that will enable us to acquire the needed data under the defined use scenarios. These include two Setups for the land-based scenario, with the first one constituting a basic platform that is mostly comprised of commercially available components, while the second one adds our wearable sweat biosensor, and it also offers a high degree of smart textile integration. Similarly, for the water-based scenario we target two Setup variants, comprising a low-risk basic version relying on in-water data communication via waveguides integrated into a full body swimsuit, while the higher risk advanced version targets in-water data communication by wireless sub-GHz transceivers.
To ease the development of these Setups we specified a list of technologies that are to be integrated into the Setups. These technologies include different wearable sensors (for measuring heart rate, acceleration, muscle activation, respiration rate, blood oxygenation levels, and various biochemical sweat parameters), the central edge-AI processor, in-water communication modules, a gateway between the central processor and the cloud environment, various textile integration techniques, and procedures for data analysis, edge-AI processing and data encryption / security. We initiated the development and technical validation of these technologies within the first project period, and we will finalize their development and integration into our demonstrator Setups during the second project period, including also the end-user validation of our demonstrator Setups.
1) Land-based scenario: An edge-AI algorithm for predicting the personal performance indicator and metabolic thresholds in real-time during training by analysing the data streams of a multitude of wearable sensors, including also biochemical data delivered by our wearable sweat biosensor. The performance indicator and metabolic thresholds delineate exercise intensity domains and provide quantitative benchmarks for training prescription, and are, therefore, equally relevant both for athletic and rehabilitation use scenarios. However, a valid assessment of these parameters can currently only be obtained in a laboratory environment by invasive methods.
2) Water-based scenario: An edge-AI algorithm that provides live assessment of swimming techniques by real-time analysis of biomechanical parameters delivered by multiple sensors placed at various critical body positions. Here, currently available systems can only record data for post-training analysis, which is a direct result of the non-existence of effective in-water data communication technology. By leveraging the in-water communication technology developed in the project, we target to deliver for the first time live swimming technique assessment, thereby enabling direct response to swimmers and swimming trainers.
With the edge-AI algorithms defined, we next specified distinct demonstrator Setups that will enable us to acquire the needed data under the defined use scenarios. These include two Setups for the land-based scenario, with the first one constituting a basic platform that is mostly comprised of commercially available components, while the second one adds our wearable sweat biosensor, and it also offers a high degree of smart textile integration. Similarly, for the water-based scenario we target two Setup variants, comprising a low-risk basic version relying on in-water data communication via waveguides integrated into a full body swimsuit, while the higher risk advanced version targets in-water data communication by wireless sub-GHz transceivers.
To ease the development of these Setups we specified a list of technologies that are to be integrated into the Setups. These technologies include different wearable sensors (for measuring heart rate, acceleration, muscle activation, respiration rate, blood oxygenation levels, and various biochemical sweat parameters), the central edge-AI processor, in-water communication modules, a gateway between the central processor and the cloud environment, various textile integration techniques, and procedures for data analysis, edge-AI processing and data encryption / security. We initiated the development and technical validation of these technologies within the first project period, and we will finalize their development and integration into our demonstrator Setups during the second project period, including also the end-user validation of our demonstrator Setups.
Of the technologies developed during the first project period, in particular the following results are currently most promising to deliver high scientific and technological impact:
Wearable sweat biosensor: It comprises an electrochemical sensor chip for multi-parameter sweat analysis and a matching microfluidic chip for sweat collection. The sensor chip contains a total of 48 sensing elements on a chip area of 7.5 x 7.5 mm2 for quantifying sweat conductivity, pH & electrolyte levels, and glucose & lactate concentrations (validated to date for in-vitro determination of conductivity, pH, and potassium / sodium ions). The sensor chip is bonded to the microfluidic chip, which has been demonstrated in-vivo to support sweat flow rates of up to 10 µl/min, with the sweat directed towards the sensor chip electrodes by a meander-shaped channel.
In-water communication technology: For wireless in-water data communication, sub-GHz transceivers have been developed and validated for long range data transmission from a submerged transmitter to an out of the water receiver (communication scenario from central processor to gateway), as well as for underwater multi-hop data communication from a transmitter to a receiver via an intermediate relay transceiver (all devices attached to the submerged leg of a swimmer, communication scenario from sensor node to central processor).
Initial version of the AI-algorithm for swimming technique assessment: Based on an open source IMU sensor database for swimming applications, an initial version of the AI-algorithm for swimming technique assessment has been developed and demonstrated.
Peripheral blood oxygen saturation (SpO2) sensor: A custom-designed SpO2 sensor prototype based on reflective mode photoplethysmography (PPG) has been developed and is currently undergoing technical validation.
Dry textile electrodes for surface electromyography (sEMG) measurements: We demonstrated similar signal quality using textile integrated dry electrodes (instead of the standardly employed pre-gelled electrodes) for recording sEMG signals, offering a route towards better integrated and more user-friendly muscle activation measurements.
Wearable sweat biosensor: It comprises an electrochemical sensor chip for multi-parameter sweat analysis and a matching microfluidic chip for sweat collection. The sensor chip contains a total of 48 sensing elements on a chip area of 7.5 x 7.5 mm2 for quantifying sweat conductivity, pH & electrolyte levels, and glucose & lactate concentrations (validated to date for in-vitro determination of conductivity, pH, and potassium / sodium ions). The sensor chip is bonded to the microfluidic chip, which has been demonstrated in-vivo to support sweat flow rates of up to 10 µl/min, with the sweat directed towards the sensor chip electrodes by a meander-shaped channel.
In-water communication technology: For wireless in-water data communication, sub-GHz transceivers have been developed and validated for long range data transmission from a submerged transmitter to an out of the water receiver (communication scenario from central processor to gateway), as well as for underwater multi-hop data communication from a transmitter to a receiver via an intermediate relay transceiver (all devices attached to the submerged leg of a swimmer, communication scenario from sensor node to central processor).
Initial version of the AI-algorithm for swimming technique assessment: Based on an open source IMU sensor database for swimming applications, an initial version of the AI-algorithm for swimming technique assessment has been developed and demonstrated.
Peripheral blood oxygen saturation (SpO2) sensor: A custom-designed SpO2 sensor prototype based on reflective mode photoplethysmography (PPG) has been developed and is currently undergoing technical validation.
Dry textile electrodes for surface electromyography (sEMG) measurements: We demonstrated similar signal quality using textile integrated dry electrodes (instead of the standardly employed pre-gelled electrodes) for recording sEMG signals, offering a route towards better integrated and more user-friendly muscle activation measurements.