Gait Event Detection and Travel Distance Using Waist-Worn Accelerometers across a Range of Speeds: Automated Approach

Albara Ah Ramli, Xin Liu, Kelly Berndt, Chen-Nee Chuah, Erica Goude, Lynea B. Kaethler, Amanda Lopez, Alina Nicorici, Corey Owens, David Rodriguez, Jane Wang, Daniel Aranki, Craig M. McDonald, Erik K. Henricson

TL;DR

This paper introduces a novel, automated approach using waist-worn accelerometers to accurately measure gait features like step count, stride length, and travel distance. Their method, which involves individual calibration and machine learning, significantly outperforms standard smartphone pedometers, offering a promising tool for remote mobility monitoring, especially for individuals with conditions like Duchenne Muscular Dystrophy (DMD).

What Problem Does This Paper Address?

Accurately assessing gait characteristics (like step count, length, and speed) is crucial for understanding mobility, particularly for individuals with progressive conditions like Duchenne Muscular Dystrophy (DMD). These individuals experience changes in gait patterns, making it difficult for standard, unsupervised methods to reliably capture their mobility. Existing solutions often rely on clinical settings or specialized equipment, limiting their use in community-based, long-term monitoring. Furthermore, consumer-grade pedometers, while convenient, often struggle with accuracy, especially at slower speeds or with atypical gaits. This paper aims to bridge this gap by developing a robust, automated system that uses readily available accelerometers to accurately measure gait features across a range of speeds and for individuals with varying mobility levels.

Methods

The core of this research is a multi-step, machine learning-based approach for gait analysis using data from a single accelerometer worn at the waist. The process involves:

  • Individualized Calibration: Recognizing that gait patterns vary significantly between individuals, the system employs a calibration phase. This involves collecting accelerometer data at five different, pre-defined speeds (SC-L1 to SC-L5).
  • Step Detection: Raw anteroposterior accelerometer signals are filtered to remove noise. Peaks in the filtered signal are identified as step events. This method is designed to be robust across different walking and running speeds.
  • Stride Length Estimation: A key innovation is the use of an individualized nonlinear regression model. This model learns the relationship between average peak acceleration values and stride lengths for each participant, based on the calibration data. This allows for accurate stride length prediction even for unseen gait activities.
  • Distance Estimation: Once step lengths and counts are determined, the total distance traveled is calculated by accumulating individual step lengths.
  • Gait Pattern Representation: The system can also generate a normalized gait cycle map, allowing for visual comparison of gait patterns between individuals and across different speeds.

The method leverages a combination of clinical observation (for initial calibration and ground truth) and machine learning for automated feature extraction.

Datasets

The study utilized data collected from:

  • 30 Participants: 15 children with Duchenne Muscular Dystrophy (DMD) and 15 typically developing (TD) controls.
  • Gait Activities: Participants performed various standardized and free-roaming gait activities, including:
  • Speed-calibration tests (SC-L1 to SC-L5) at different walking/running speeds.
  • 6-minute walk test (6 MWT).
  • 100-meter run/walk (100 MRW).
  • Free-walk (FW) within a building.
  • Ground Truth: Data was compared against ground-truth observations derived from video recordings.

No new benchmark datasets were introduced; instead, the study focused on evaluating their method against existing ground truth and a common consumer device (iPhone pedometer).

Key Insights and Ideas

  • Individualized Calibration is Key: The paper strongly emphasizes that a one-size-fits-all approach to gait analysis with accelerometers is insufficient. Tailoring the model to each individual's unique gait patterns through calibration is critical for accuracy.
  • Waist-Worn Accelerometer Sufficiency: The research demonstrates that a single accelerometer placed at the waist, near the body's center of mass, is sufficient for accurate gait event detection and distance estimation, eliminating the need for more complex multi-sensor setups.
  • Machine Learning for Atypical Gaits: The ML-based approach proves effective in handling the varied and often atypical gait patterns observed in individuals with DMD, outperforming simpler methods.
  • Beyond Step Counting: While pedometers focus on step counts, this method extends to stride length and distance, providing a more comprehensive picture of mobility.

Key Results

The proposed Walk4Me system demonstrated high accuracy across various gait activities and speeds:

  • Strong Correlation: Model predictions for step counts, distance traveled, and step length showed a strong correlation with ground-truth observations (Pearson's r ranging from -0.9929 to 0.9986, p < 0.0001).
  • Low Error Rates:
    • Step Counts: Mean (SD) percentage error of 1.49% (7.04%) for combined 6 MWT, 100 MRW, and FW tasks.
  • Distance Traveled: Mean (SD) percentage error of 1.18% (9.91%) for combined 6 MWT, 100 MRW, and FW tasks.
  • Step Length: Mean (SD) percentage error of 0.37% (7.52%) for combined 6 MWT, 100 MRW, and FW tasks.
  • Superior to Pedometer: The Walk4Me system significantly outperformed the iPhone's built-in pedometer, which exhibited much higher error rates (e.g., average step count error of 48.46% vs. 3.46% for Walk4Me).
  • No Significant Difference Between Groups: The mean error percentages for step counts, distance, and step length were not statistically different between DMD participants and TD controls, indicating the method's effectiveness across different mobility levels.

Limitations and Future Work

  • Limitations:
    • Calibration Requirement: The method requires an initial calibration phase, which involves a trained clinical evaluator. While short, it's still a step beyond fully unsupervised methods.
  • Specific Speeds for Calibration: The calibration uses specific speed points (SC-L1 to SC-L5). While effective, the range and granularity of these points might be a limitation for individuals with extremely limited or highly variable gait speeds.
  • Data Availability Statement: The paper notes that due to privacy concerns, raw data will only be shared under specific conditions, which might limit external replication without further agreements.
  • Future Work:
    • Fully Unsupervised Calibration: The authors suggest extending the method to achieve fully unsupervised calibration, removing the need for a trained observer.
  • Broader Range of Activities: Applying the method to a wider array of community-based activities and longer-term monitoring.
  • Integration with Other Sensors: Exploring the integration of this method with other wearable sensors for a more comprehensive assessment of mobility and health.
  • Clinical Validation: Further validation in larger, more diverse clinical populations to solidify its utility in clinical decision-making.

Summary updated

Abstract

Estimation of temporospatial clinical features of gait (CFs), such as step count and length, step duration, step frequency, gait speed, and distance traveled, is an important component of community-based mobility evaluation using wearable accelerometers. However, accurate unsupervised computerized measurement of CFs of individuals with Duchenne muscular dystrophy (DMD) who have progressive loss of ambulatory mobility is difficult due to differences in patterns and magnitudes of acceleration across their range of attainable gait velocities. This paper proposes a novel calibration method. It aims to detect steps, estimate stride lengths, and determine travel distance. The approach involves a combination of clinical observation, machine-learning-based step detection, and regression-based stride length prediction. The method demonstrates high accuracy in children with DMD and typically developing controls (TDs) regardless of the participant’s level of ability. Fifteen children with DMD and fifteen TDs underwent supervised clinical testing across a range of gait speeds using 10 m or 25 m run/walk (10 MRW, 25 MRW), 100 m run/walk (100 MRW), 6-min walk (6 MWT), and free-walk (FW) evaluations while wearing a mobile-phone-based accelerometer at the waist near the body’s center of mass. Following calibration by a trained clinical evaluator, CFs were extracted from the accelerometer data using a multi-step machine-learning-based process and the results were compared to ground-truth observation data. Model predictions vs. observed values for step counts, distance traveled, and step length showed a strong correlation (Pearson’s r = −0.9929 to 0.9986, p < 0.0001). The estimates demonstrated a mean (SD) percentage error of 1.49% (7.04%) for step counts, 1.18% (9.91%) for distance traveled, and 0.37% (7.52%) for step length compared to ground-truth observations for the combined 6 MWT, 100 MRW, and FW tasks. Our study findings indicate that a single waist-worn accelerometer calibrated to an individual’s stride characteristics using our methods accurately measures CFs and estimates travel distances across a common range of gait speeds in both DMD and TD peers.
BibTeX
@inproceedings{ramli2024gait,
  author = {Albara Ah Ramli and Xin Liu and Kelly Berndt and Chen-Nee Chuah and Erica Goude and Lynea B. Kaethler and Amanda Lopez and Alina Nicorici and Corey Owens and David Rodriguez and Jane Wang and Daniel Aranki and Craig M. McDonald and Erik K. Henricson},
  title = {{Gait Event Detection and Travel Distance Using Waist-Worn Accelerometers across a Range of Speeds: Automated Approach}},
  booktitle = {Sensors},
  year = {2024},
  url = {https://openreview.net/forum?id=9HYXk5CUBS},
}
Figure