In-sensor human gait analysis with machine learning in a wearable microfabricated accelerometer

Dion, Guillaume; Tessier-Poirier, Albert; Chiasson-Poirier, Laurent; Morissette, Jean-François; Brassard, Guillaume; Haman, Anthony; Turcot, Katia; Sylvestre, Julien

doi:10.1038/s44172-024-00193-5

Cited by 3 publications

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

In-sensor human gait analysis with machine learning in a wearable microfabricated accelerometer

Dion,

Tessier-Poirier,

Chiasson-Poirier

et al. 2024

Commun Eng

View full text Add to dashboard Cite

In-sensor computing could become a fundamentally new approach to the deployment of machine learning in small devices that must operate securely with limited energy resources, such as wearable medical devices and devices for the Internet of Things. Progress in this field has been slowed by the difficulty to find appropriate computing devices that operate using physical degrees of freedom that can be coupled directly to degrees of freedom that perform sensing. Here we leverage reservoir computing as a natural framework to do machine learning with the degrees of freedom of a physical system, to show that a micro-electromechanical system can implement computing and the sensing of accelerations by coupling the displacement of suspended microstructures. We present a complete wearable system that can be attached to the foot to identify the gait patterns of human subjects in real-time. The computing efficiency and the power consumption of this in-sensor computing system is then compared to a conventional system with a separate sensor and digital computer. For similar computing capabilities, a much better power efficiency can be expected for the highly-integrated in-sensor computing devices, thus providing a path for the ubiquitous deployment of machine learning in edge computing devices.

show abstract

In-sensor human gait analysis with machine learning in a wearable microfabricated accelerometer

Dion,

Tessier-Poirier,

Chiasson-Poirier

et al. 2024

Commun Eng

View full text Add to dashboard Cite

show abstract

Electrothermal Tunable MEMS Oscillators for MEMS-Based Reservoir Computing

Lee,

Wang,

Chuang

et al. 2024

IEEE Sens. Lett.

View full text Add to dashboard Cite

Design of an Integrated Model for Gait Identification Using DDPG, Sparse Group Lasso, and Stacked Generalization

Urkude,

Misal,

Choubey

et al. 2024

Preprint

View full text Add to dashboard Cite

Applications in security, healthcare, and human-computer interaction critically require accurate gait identification under complex environmental conditions such as varying lighting and background noise. Current approaches are usually unable to adapt to dynamic, highdimensional environments, with reduced accuracy of feature extraction and classification. This paper bridges the gap by offering an overview of a multi-stage framework that merges the advanced techniques of machine learning with those of reinforcement learning for preemptive optimization. It begins by using Deep Deterministic Policy Gradient for a preprocessing module: environmental parameters are dynamically adjusted so that their real-time data quality is optimized. The module is then followed by a phase in multi-domain feature extraction using Sparse Group Lasso along with KMeans clustering, thereby improving representativeness while reducing dimensionality by 50–60%. We have used a hybrid of stacked generalization, in this case of XGBoost and LightGBM, because this provides a better overall classification accuracy. Refined temporal post-processing at the hidden Markov model and Auto-Regressive Integrated Moving Average (ARIMA) results in enhanced phase transitions that may be gait-based, thus improving the identification accuracy. As the final step, we use Proximal Policy Optimization to implement feedback-driven reinforcement learning, where improvements are incrementally made by updating the model with iterative feedback. This new method enhances the correctness of feature extraction by 12% in complex environments. Overall classification accuracy increases by 5–6% and reaches 95%. False positives in gait phase transitions decrease as well, further increasing the system robustness and reliability in real-world applications.

show abstract

In-sensor human gait analysis with machine learning in a wearable microfabricated accelerometer

Cited by 3 publications

References 48 publications

In-sensor human gait analysis with machine learning in a wearable microfabricated accelerometer

In-sensor human gait analysis with machine learning in a wearable microfabricated accelerometer

Electrothermal Tunable MEMS Oscillators for MEMS-Based Reservoir Computing

Design of an Integrated Model for Gait Identification Using DDPG, Sparse Group Lasso, and Stacked Generalization

Contact Info

Product

Resources

About