Related Experiment Video
Updated: Jan 16, 2026
Quantified Assessment of Infant's Gross Motor Abilities Using a Multisensor Wearable
Published on: May 17, 2024
Gait phase recognition of children with cerebral palsy via deep learning based on IMU data from a soft ankle
Zhi Pang1, Zewei Li2, Ying Li3
1School of Health Science and Engineering, University of Shanghai for Science and Technology, Shanghai, China.
Insights
A new deep learning model, SDA-LSTM, accurately identifies gait phases in children with Cerebral Palsy (CP). This technology shows promise for improving rehabilitation and assistive devices for individuals with CP.
Area of Science:
- Biomedical Engineering
- Rehabilitation Technology
- Artificial Intelligence in Medicine
Background:
- Accurate gait-phase identification is crucial for effective rehabilitation in children with Cerebral Palsy (CP).
- Existing methods may struggle with the complex and noisy gait patterns typical of CP patients in natural environments.
Purpose of the Study:
- To develop and validate a deep learning framework for precise gait-phase identification in children with CP.
- To assess the robustness of the proposed model under noisy conditions.
Main Methods:
- A novel SDA-LSTM (stacked denoising autoencoder with long short-term memory network) deep learning framework was proposed.
- Data from ankle-mounted IMUs and plantar-pressure insoles were synchronized for six children with mild CP.
- The SDA layer extracted features, and the LSTM module modeled temporal dependencies for gait phase classification.
Main Results:
- The SDA-LSTM framework achieved 97.83% accuracy in noise-free conditions, outperforming SVM, random forest, and standalone LSTM.
- The model demonstrated robust performance under additive Gaussian noise, maintaining 90.96% accuracy at 10 dB SNR.
- The framework effectively handled complex and heterogeneous gait patterns.
Conclusions:
- The SDA-LSTM model offers a highly accurate and robust solution for gait-phase identification in children with CP.
- This technology has significant translational potential for clinical gait assessment and integration with exoskeletal assistance systems.
Abstract:
Accurate gait-phase identification in children with Cerebral Palsy (CP) constitutes a pivotal prerequisite for evidence-based rehabilitation. Addressing the precise detection of gait disturbances under natural ambulation, we propose a deep-learning framework that integrates a stacked denoising autoencoder (SDA) with a long short-term memory network (SDA-LSTM) to classify four canonical gait phases. A community-oriented dataset was constructed by synchronizing ankle-mounted inertial measurement units (IMU) with plantar-pressure insoles; natural gait sequences of six children with mild CP were acquired in open environments. The SDA layer robustly extracts discriminative representations from non-stationary, high-noise signals, whereas the LSTM module models inter-phase temporal dependencies, thereby enhancing generalization cross-user. In noise-free conditions the SDA-LSTM framework attained 97.83% accuracy, significantly exceeding SVM (94.68%), random forest (96.05%), and standalone LSTM (95.86%). Under additive Gaussian noise with SNR ranging from 5 to 30 dB, the model preserved stable performance; at 10 dB SNR (Signal-to-Noise Ratio), accuracy remained 90.96%, corroborating its exceptional robustness. These findings demonstrate that SDA-LSTM effectively handles the complex, heterogeneous gait patterns of children with CP and is readily deployable for clinical assessment and exoskeletal assistance systems, indicating substantial translational potential.
