Wrist-Worn Pedometer Step Detection Using Adaptive Peak Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional pedometers inaccurately count steps when worn on the wrist due to the combination of arm sway and torso motion, which complicates the detection of step acceleration and leads to inaccurate step counts.
Innovation Solution
A 3-axis accelerometer-based pedometer system that uses a polynomial fit filter and adaptive peak detection algorithm to differentiate between arm sway and step motion, allowing for accurate step counting regardless of the device's orientation, and includes a stride length model to estimate distance traveled.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a simple accelerometer-based detection method is used, then device complexity is reduced, but measurement precision deteriorates due to inability to differentiate arm sway from step motion
Solution Approach 1:
The detection algorithm segments the acceleration signal into distinct components by identifying peaks and valleys separately. Each peak represents potential step motion while valleys represent arm sway or resting state. This segmentation allows the system to count only meaningful peaks as steps, improving accuracy without requiring complex processing.
Solution Approach 2:
The system dynamically adjusts detection thresholds based on the observed motion patterns. By continuously monitoring the acceleration signal and adapting peak/valley detection criteria to current motion characteristics, the algorithm maintains high accuracy across varying walking speeds and arm swing intensities without manual recalibration.
2Ease of operation
If the pedometer is worn on the wrist, then ease of operation and wearability are improved, but measurement precision deteriorates due to arm sway interfering with step detection
Solution Approach 1:
The algorithm converts the harmful effect of arm sway into a useful detection mechanism. By identifying valleys in the acceleration signal as indicators of arm sway or resting state, the system uses this information to suppress false step counts. The harmful arm motion pattern becomes the basis for distinguishing true steps from false detections.
Solution Approach 2:
The detection algorithm acts as an intermediary between the raw acceleration data and the final step count. It processes the signal through peak and valley identification, filtering out noise and spurious signals before generating the final measurement. This intermediary processing layer isolates true step motion from interfering arm sway movements.
3Measurement precision
If adaptive peak detection with polynomial fit filter is applied, then measurement precision is improved for differentiating step motion from arm sway, but device complexity increases
Solution Approach 1:
The system applies a polynomial fit filter with specific order (partial action) to smooth the acceleration signal and reduce noise. This partial filtering approach provides sufficient precision for step detection without over-processing the signal. The filter order and parameters are optimized to achieve the minimum necessary processing for accurate detection, avoiding excessive computational complexity.
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
The system provides accurate step counting and distance measurement when worn on the wrist, matching the accuracy of belt-worn pedometers, with reduced errors and improved reliability across various walking conditions.
Implementation Method 1
A 3-axis accelerometer-based pedometer system that uses a polynomial fit filter and adaptive peak detection algorithm to differentiate between arm sway and step motion
Data Source
AI summary
A pedometer with a three-axis accelerometer provides reliable step counts while worn on the wrist. Three-axis accelerometer data is combined into a single combined data stream. Each positive slope region around an inflection point in the combined data stream that has positive slope, a magnitude that exceeds an amplitude threshold value and that spans a time period that exceeds a time threshold value is identified. Each negative slope region around an inflection point in the combined data stream that has negative slope, a magnitude that exceeds an amplitude threshold value and that spans a time period that exceeds a time threshold value is identified. A step count is incremented for each occurrence of an identified positive slope region that is separated by an identified negative slope region as a step.