article · Sensors
Micro Electronic Mechanical Systems sensors embedded in smartphones and wearable devices, including accelerometers, gyroscopes, and magnetometers, enable automated human activity recognition. In this research, a dataset covering nine everyday movements was compiled, comprising laying down, remaining stationary, walking, brisk walking, running, climbing stairs, descending stairs, squatting, and cycling. Multiple standard machine learning models were trained on the sensor data, including decision trees, k-nearest neighbours, multinomial logistic regression, Gaussian naive Bayes, support vector machines, and random forest. Among these standard approaches, the random forest classifier achieved the highest accuracy at 95 percent. In addition, a custom bidirectional long short-term memory recurrent neural network was developed, achieving an improved test accuracy of 98.1 percent. This artificial intelligence methodology outperforms the traditional algorithmic techniques currently utilised in wearable technologies.
Accurate human activity recognition is vital for personal fitness trackers, smartwatches, and health-monitoring systems. By applying advanced recurrent neural networks to data from standard built-in motion sensors, smart devices can distinguish subtle differences between daily physical activities with greater precision. This can improve the reliability of automated movement tracking without requiring dedicated or costly new hardware.
The work represents applied and tested research suitable for integration into consumer wearables, smartphones, and mobile health tracking applications. Hardware manufacturers and software developers building fitness or mobility platforms are the primary potential users. While the high classification accuracy is demonstrated on experimental sensor datasets, moving the models into commercial products would require testing and adaptation for real-time processing on embedded device software.
AI-generated from the published abstract. Always read the original work before citing.
The integration of Micro Electronic Mechanical Systems (MEMS) sensor technology in smartphones has greatly improved the capability for Human Activity Recognition (HAR). By utilizing Machine Learning (ML) techniques and data from these sensors, various human motion activities can be classified. This study performed experiments and compiled a large dataset of nine daily activities, including Laying Down, Stationary, Walking, Brisk Walking, Running, Stairs-Up, Stairs-Down, Squatting, and Cycling. Several ML models, such as Decision Tree Classifier, Random Forest Classifier, K Neighbors Classifier, Multinomial Logistic Regression, Gaussian Naive Bayes, and Support Vector Machine, were trained on sensor data collected from accelerometer, gyroscope, and magnetometer embedded in smartphones and wearable devices. The highest test accuracy of 95% was achieved using the random forest algorithm. Additionally, a custom-built Bidirectional Long-Short-Term Memory (Bi-LSTM) model, a type of Recurrent Neural Network (RNN), was proposed and yielded an improved test accuracy of 98.1%. This approach differs from traditional algorithmic-based human activity detection used in current wearable technologies, resulting in improved accuracy.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.3390/s23031275
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
