Half Step Frequency Motion Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for classifying motion contexts using tri-axial accelerometers in mobile devices are insufficient in distinguishing between walking/running and other motions like fiddling or shaking, as they fail to accurately detect the half-step frequency relationship in acceleration data.
Innovation Solution
The method involves collecting accelerometer data over a selected time window and analyzing it in both time and frequency domains to detect the presence or absence of a half-step frequency relationship between the x, y, and z axes, using techniques such as Fast Fourier Transform, auto-correlation, and peak analysis to determine the motion state, thereby distinguishing between walking/running and other activities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional motion classification methods are used based on peak detection at step frequency and half step frequency, then the classification can be performed using simple peak analysis, but the method fails to reliably distinguish between walking/running and other motions like fiddling or shaking
Solution Approach 1:
The patent segments the motion analysis into multiple frequency components by performing Fast Fourier Transform on accelerometer data from multiple axes. Instead of relying on a single peak detection, the method divides the frequency spectrum into step frequency peaks and half-step frequency peaks, analyzing their relationships separately to improve classification reliability
Solution Approach 2:
The patent transitions from time-domain peak detection to frequency-domain analysis using Fast Fourier Transform. This dimensional change allows the system to detect the half-step frequency relationship more reliably by examining spectral components across multiple axes (x, y, z) rather than relying on temporal patterns alone
2Measurement precision
If the system analyzes accelerometer data in both time and frequency domains to detect half-step frequency relationships, then motion classification accuracy improves, but the computational complexity and processing time increase
Solution Approach 1:
The patent uses the Fast Fourier Transform algorithm to simultaneously extract multiple frequency components (step frequency peaks, half-step frequency peaks, and their relationships) from the same accelerometer data. This multi-functional approach allows the system to detect various motion characteristics including walking, running, fiddling, and shaking using a single transformation method
Solution Approach 2:
The system uses the inherent spectral relationships in the accelerometer data itself to classify motion states. By detecting the half-step frequency relationship that naturally exists in walking and running motions, the method allows the data to self-reveal its classification category without requiring external reference data or complex machine learning models
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
Disclosed is an apparatus and method for classifying a motion state of a mobile device. In one embodiment, accelerometer data representing acceleration components along orthogonal x, y, and z axes of the mobile device are collected. A presence or absence of a half-step frequency relationship between the accelerometer data is determined. Last, the motion state of the device is determined based at least in part on the presence or absence of the half-step frequency relationship.