Physical State Estimation Using Multi-Source Movement Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating the physical state of a movable object, such as a mobile phone, using inertial measurement unit (IMU) data are limited in reliability due to the reliance on a single data source, which can lead to inaccurate classification of movement states.
Innovation Solution
An apparatus and method that combine probability mass functions from multiple information sources, including IMU data and additional information, to select the most probable movement class and estimate the physical state, with a quality measure indicating the reliability of each source to adjust the accuracy of the estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If only IMU sensor data is used for movement state classification, then the device complexity is reduced, but the reliability of physical state estimation deteriorates
Solution Approach 1:
The patent combines probability mass functions from multiple information sources including IMU sensor data, wireless signal characteristics, and environmental data to estimate the physical state of a movable object. This merging of multiple data sources improves reliability by compensating for the limitations of any single source, while the probabilistic framework provides a systematic way to integrate heterogeneous information without exponentially increasing system complexity.
2Measurement precision
If multiple probability mass functions from different information sources are combined, then the measurement precision of movement class classification is improved, but the device complexity increases
Solution Approach 1:
The patent transforms multiple different types of information sources into a unified probabilistic parameter representation (probability mass functions over movement classes). This parameter transformation allows heterogeneous data to be compared and combined on a common scale, improving classification precision while maintaining manageable complexity through the consistent probabilistic framework.
Solution Approach 2:
The probability mass function serves as an intermediary representation that mediates between raw sensor data and final movement state classification. Each information source is converted to a PMF, which then can be combined systematically. This intermediary step simplifies the integration process by providing a common language for combining diverse data sources without directly processing their raw complexity.
3Measurement precision
If a single movement model is used for physical state estimation, then the processing speed is maintained, but the accuracy of physical state estimation deteriorates
Solution Approach 1:
The patent performs preliminary classification of movement classes using combined probability mass functions from multiple information sources before selecting the appropriate movement model. This preliminary action identifies the most likely movement class, which then guides the selection of a suitable movement model. By preparing the classification result in advance, the system enables faster and more accurate model selection without requiring extensive real-time computation during the estimation phase.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
An apparatus for estimating a physical state of a movable object comprises a processor receiving or determining a probability mass function including probabilities for each of a first group of at least two movement classes, wherein the movement models of the first group being determined using sensor data from the inertial measurement unit. The processor receives at least one additional probability mass function associated with a second group of at least two movement classes, wherein the additional probability mass function has been obtained using additional information different from the sensor data. The processor combines the probability mass function and the at least one additional probability mass function to obtain a combined probability mass function over the movement classes of the first group and the second group, selects a movement class having the highest probability from the combined probability mass function, and estimates the physical state of the movable object using a movement model of the selected movement class. Each movement class is either a movement state or a movement model.