Related Experiment Video
Updated: Jul 18, 2026
06:49
Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment
Published on: December 11, 2015
9.0K
A novel smartphone-based activity recognition modeling method for tracked equipment in forest operations
Ryer M Becker1, Robert F Keefe1
1Department of Forest, Rangeland and Fire Sciences, College of Natural Resources, University of Idaho, Moscow, Idaho, United States of America.
Plos One
|April 6, 2022
Summary
Smartphone sensors can accurately recognize activities in forest management, using inertial measurement units (IMUs) and sound. This technology offers automated time studies for mechanized operations, improving efficiency and data collection.
Area of Science:
- Forestry
- Mechanical Engineering
- Data Science
Background:
- Activity recognition using smartphone Inertial Measurement Units (IMUs) is an underutilized resource for assessing work efficiency in natural resource management.
- Excavator-based mastication equipment is crucial for Ponderosa pine plantation management, but its operational efficiency is not well-quantified.
Purpose of the Study:
- To develop and validate a smartphone-based activity recognition system for excavator-based mastication equipment.
- To determine the optimal sensor data sampling frequencies, window widths, and overlap for accurate activity classification.
- To evaluate the performance of a Random Forest machine learning model for classifying specific work elements.
Main Methods:
- Collected sensor data (gyroscopes, accelerometers, decibel meters) from smartphones during mastication treatments at 10, 20, and 50 Hz.
- Extracted 9 time-domain features using sliding windows with widths of 1, 5, 7.5, and 10 seconds, and 50% or 90% overlap.
- Trained and evaluated Random Forest models with 40 parameter combinations to classify 5 work elements: masticate, clear, move, travel, and delay.
Main Results:
- The best model (50 Hz, 10-second window, 90% overlap) achieved high performance metrics: AUC (95.0% - 99.9%), sensitivity (74.9% - 95.6%), specificity (90.8% - 99.9%), precision (81.1% - 98.3%), F1-score (81.9% - 96.9%), and balanced accuracy (87.4% - 97.7%).
- Smartphone sensors effectively characterized individual work elements of mechanical fuel treatments.
- This is the first study to develop a smartphone-based activity recognition model for ground-based forest equipment.
Conclusions:
- Smartphone sensors provide an effective means to characterize work elements in mechanized forest operations.
- This technology can support land managers and operators with ubiquitous, manufacturer-independent systems for automated time studies and production analysis.
- Further development and dissemination of these models can enhance efficiency and data-driven decision-making in forest management.
