Fitness Tracking System Using Multi-Device Sensor Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current fitness tracking systems face challenges in accurately distinguishing between similar exercise activities, such as bench press with a barbell and dumbbells, due to similar physiological motion characteristics, leading to reduced accuracy in exercise predictions.
Innovation Solution
The system employs a combination of wearable devices, including smart watches and mobile phones, to generate exercise predictions using machine learning models based on time-series sensor data, oscillating signal profiles, and environment data, providing increasingly granular predictions by buffering and combining data from multiple devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single wearable device is used for exercise tracking, then device simplicity is maintained, but measurement precision deteriorates due to inability to distinguish similar exercises
Solution Approach 1:
The patent combines data from multiple wearable devices (smartwatch, mobile phone, audio device) to improve exercise prediction accuracy. By merging sensor data from different devices worn on different body parts, the system can distinguish between similar exercises that a single device cannot differentiate, directly resolving the contradiction between measurement precision and device complexity.
Solution Approach 2:
The system segments the monitoring function across multiple devices, with each device collecting specific sensor data from its location on the user's body. This segmentation allows the system to capture diverse motion patterns from different anatomical positions, improving exercise differentiation while maintaining individual device simplicity.
2Measurement precision
If multiple wearable devices are combined for data collection, then measurement precision improves through multiple data sources, but device complexity increases
Solution Approach 1:
The patent introduces a server as an intermediary that receives, processes, and integrates sensor data from multiple wearable devices. This intermediary handles the complex tasks of data synchronization, feature extraction, and exercise classification, allowing individual devices to remain simple while achieving high measurement precision through coordinated multi-device operation.
Solution Approach 2:
The server performs multiple functions including data collection, preprocessing, feature extraction, exercise classification, and feedback generation. This multi-functional approach consolidates the complexity into a single processing platform, enabling accurate exercise differentiation without requiring each wearable device to be complex.
3Productivity
If buffered sensor data is processed in real-time, then productivity of feedback is improved, but use of energy increases
Solution Approach 1:
The system performs preliminary processing of sensor data by buffering it locally on wearable devices before transmission to the server. This preliminary action includes basic filtering and feature extraction, which reduces the computational burden during real-time processing and lowers energy consumption while maintaining fast feedback response.
Solution Approach 2:
The system processes buffered sensor data at periodic intervals rather than continuously analyzing every data point. This periodic processing approach maintains real-time feedback capability while significantly reducing energy consumption compared to continuous real-time analysis, as the system can batch process data without losing temporal relevance.
Data Source
AI summary
Fitness tracking devices and methods of operating the same. The fitness tracking device includes a sensor circuit to generate sensor data; a processor coupled to the sensor circuit; and a memory coupled to the processor and storing processor-executable instructions that, when executed, configure the processor to: buffer sensor data associated with motion of the user limb; generate an exercise prediction based on a prediction model and the sensor data, the prediction model defined by one or more oscillating signal profiles to identify genus predictions for respective limb movement types about at least one sensor axis, wherein the exercise prediction is generated based on a combination of an identified genus prediction associated with the generated sensor data and environment data associated with motion of the user limb; and transmit a signal representing the exercise prediction for display on a user interface.