Micro-AI Overtraining Detection via Latent Data Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional AI systems are unsuitable for mobile devices used by athletes due to high memory and processing requirements, and the need for internet connectivity, making it difficult to detect overtraining and injury conditions effectively.
Innovation Solution
A system utilizing bifurcated data organized into latent and current data, with a microprocessor in a wearable fitness tracker that can analyze data independently of internet connectivity to provide alerts for overtraining, using a neural network or other classifiers to determine overtraining conditions based on heart rate and other features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional AI techniques are used to analyze training data and detect overtraining conditions, then the accuracy of injury prediction is improved, but the memory requirements and processing capability become prohibitively large for mobile devices
Solution Approach 1:
The patent segments the AI system into two parts: a training phase that occurs on powerful servers using large datasets, and an inference phase that occurs on the mobile device using a pre-trained model. The large dataset is divided into training data (for creating the model) and latent data (stored on the device for real-time analysis). This segmentation allows the mobile device to use AI without storing the entire original dataset.
Solution Approach 2:
The system performs preliminary action by pre-training the AI model on servers using extensive historical data before deploying it to the mobile device. The model is prepared in advance with all the knowledge it needs from large datasets, so that during actual use on the device, it can make accurate predictions without requiring access to the original large training data.
2Reliability
If conventional AI techniques are used to provide real-time overtraining analysis, then the reliability of injury warning is improved, but the need for internet connectivity and data transmission increases
Solution Approach 1:
The patent segments data into two categories: latent data (historical training data stored on the device) and current data (real-time sensor data). The latent data is updated periodically when internet is available, but the system can operate independently using only the stored latent data and current sensor data, eliminating the need for continuous internet connectivity.
Solution Approach 2:
The mobile device serves itself by storing essential historical data locally and performing AI analysis independently without requiring external servers or internet connectivity during operation. The device maintains its own latent data repository and can autonomously analyze current training data against historical patterns to provide injury warnings.
3Loss of information
If large datasets are transmitted via cellular coverage for AI analysis, then the comprehensiveness of training data is improved, but the cost and practicality become prohibitive
Solution Approach 1:
The patent extracts only the essential elements of the AI system from the server environment and places them on the mobile device. Instead of transmitting large datasets for every analysis, the system extracts the trained model parameters and essential historical patterns into a compact form that can be stored locally and used indefinitely without further data transmission.
Solution Approach 2:
The system changes the parameters of the AI implementation by using a pre-trained model with fixed weights and parameters on the mobile device, rather than attempting to run full training algorithms. This parameter transformation allows the device to perform accurate analysis with minimal computational resources and no need for large data transmissions.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables efficient identification and warning of overtraining conditions on mobile devices, reducing memory and processing needs, allowing for real-time alerts without relying on continuous internet access, thereby preventing injuries.
Implementation Method 1
The fitness tracker includes a heart rate sensor
Data Source
AI summary
A system 300 and method for biologically monitoring the fitness of an athlete, and providing a warning 338 when an overtraining condition is determined in order to reduce injury. Through implementation of an efficient system architecture, micro-artificial intelligence use is practical for mobile situations where internet coverage is deficient or non-existent.