Sensor Fusion for Activity Classification Using State Machines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for classifying user activity using sensors, such as those embedded in wearable devices, face inaccuracies due to the limitations of single-sensor data, particularly in differentiating between similar activities and orientations, and suffer from noise and discomfort when multiple sensors are worn.
Innovation Solution
A method that combines data from multiple sensors, including a wrist-worn device and head-worn device, to improve detection and classification accuracy by preprocessing and fusing sensor data, using a state machine to correct misclassifications, and leveraging contextual information for more robust activity logging.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple sensors are worn on different body parts to improve activity classification accuracy, then measurement precision is improved, but device complexity and user discomfort increase
Solution Approach 1:
The patent combines data from multiple sensors worn at different body locations (wrist, head, etc.) to improve activity classification accuracy. By fusing sensor data through a state machine that processes inputs from multiple sources, the system achieves more accurate activity detection without requiring each individual sensor to be perfectly accurate, thus resolving the contradiction between using multiple sensors and system complexity.
Solution Approach 2:
The state machine processor serves multiple functions: it processes sensor data from various body locations, performs activity classification, and handles data fusion. This multi-functional approach reduces the need for separate dedicated systems for each sensing location, thereby improving measurement precision while controlling overall device complexity.
2Measurement precision
If multiple sensors are worn on different body parts to improve activity classification accuracy, then measurement precision is improved, but ease of operation deteriorates due to user discomfort
Solution Approach 1:
The system merges data from multiple sensors worn at different body parts (wrist, head, etc.) to achieve accurate activity classification. By processing combined sensor inputs through a state machine, the system obtains superior measurement precision while the user only needs to wear a few common devices, maintaining ease of operation.
Solution Approach 2:
The patent utilizes sensors embedded in universally worn devices such as smartwatches and headphones, which serve multiple functions (notification, audio, health tracking). This approach improves activity classification accuracy by leveraging data from these multi-functional devices without requiring users to wear additional specialized sensors, thus maintaining ease of operation.
3Ease of operation
If sensors are placed on body parts far from the center of mass to enable wearable device placement, then ease of operation is improved, but measurement precision deteriorates due to motion noise
Solution Approach 1:
The system combines sensor data from multiple body locations, including peripheral locations like the wrist that are convenient for device placement. By fusing data from these convenient locations with sensors at other body parts, the system maintains ease of operation while achieving accurate motion detection through data fusion that compensates for the noise inherent in peripheral sensor placements.
Solution Approach 2:
The state machine acts as an intermediary that processes and reconciles sensor data from convenient but noisy locations (like the wrist) with data from other sensors. It filters and integrates the information to produce accurate activity classification, effectively mediating between the convenience of peripheral placement and the need for precise motion detection.
4Ease of operation
If sensors embedded in common wearable devices are used to improve ease of operation, then ease of operation is improved, but measurement precision deteriorates due to limited sensor locations
Solution Approach 1:
The patent merges data from sensors embedded in common wearable devices worn at different body locations (wrist-worn smartwatch, head-worn headphones, etc.). By combining the information from these conveniently worn devices through a state machine processor, the system achieves accurate activity detection and classification while maintaining ease of operation through the use of familiar, comfortable devices.
Solution Approach 2:
The system leverages sensors in universally worn devices like smartwatches and headphones that serve multiple functions (communication, audio, health monitoring). By utilizing these multi-functional devices already integrated into users' daily routines, the system improves ease of operation while achieving accurate activity detection through the combined data from their various sensors.
Data Source
AI summary
In one embodiment, a method includes accessing first sensor data from a first sensor worn on a first portion of a user's body and accessing second sensor data from a second sensor worn on a second portion of the user's body. The method includes determining, based on both the first sensor data and the second sensor data, one or more first features related to the user's activity and determining, based on the first features, an initial classification of the user's activity. When the initial classification indicates a class that includes one or more subclasses that are more distinguishable by one of the sensors, then a specific subclassification may be determined based on sensor data from only that one sensor. Otherwise, the classification of the user's activity may be based on the one or more first features that use data from both sensors.