An existing Blautic product, Ziven Active, uses electronic wearables with ECG/EMG and 6-axis movement sensors along with smart textile garments to gather data during physical exercises.
The professional generates personalized datasets by repeating the target exercise to train ML models. The exercise can be retrained until optimal results are achieved.
These models can be assigned by the professional to their users/patients through the APP to make sure they carry out the exercises being monitored and receiving feedback in real time. The result of each exercise is reported back to the professionals in order to control all the treatment/improvement plan.
With this project we will optimize the existing generic models, improve the feedback to the end user by offering additional data regarding those input features that penalize the final result of the inference and we will add parameters like HRV and arrhythmia analysis with Lorentz Plot to give more value to the professional.
- Optimize already existing generic models that are trained by professionals
- Improve the feedback to the end user by adding eXplainable AI with valuable information regarding input features impacting the online inference results.
- Add arrhythmia AI analysis with Lorentz Plot model
- Define new data-driven revenue models
The experiment aims to provide professionals with a completely customizable tool to control their patients/users during the execution of the planned exercises.
We already forget about preprogrammed exercises and we allow each professional to train their own exercises collecting muscle activity and movement data to feed machine learning models.
It is a step to make the trained models shareable between professionals, advancing in business models regarding data services.
It is scientifically proven that well-performed physical exercise is the key to achieving a healthier society and less demanding of socio-health care. The use of models in a dynamic methodology can modify the way physical trainers and physiotherapists work, specializing in the activity of preparing exercises that can later be reused by other professionals. This type of tasks can be done for many areas of real life: wellbeing, ergonomics at work, games, virtual world.
The expected KERs are directly aligned to the company's current business as it has got products in the market within the world of health, wellness and sports. Therefore we will be able to exploit the optimized trainable models, higher added value in the information offered to professionals and users during the execution of the exercises and new important cardiology data offered to professionals being collected as the exercises are carried out.
- Percentage of improvement in accuracy of new system referred to current system for Blautic with a validation dataset
- Percentage of success on feature effect detection for validation dataset with known fixed data inputs
- Accuracy of detection of probable anomaly due to arrhythmia
- Definition of new revenue methods for the company based in data-based