Related Experiment Video
Updated: Jul 10, 2025
08:05
Design and Analysis for Fall Detection System Simplification
Published on: April 6, 2020
10.7K
Multimodal Gait Abnormality Recognition Using a Convolutional Neural Network-Bidirectional Long Short-Term Memory
Jing Li1,2, Weisheng Liang1, Xiyan Yin1
1School of Mechanical Engineering and Hubei Modern Manufacturing Quality Engineering Key Laboratory, Hubei University of Technology, Wuhan 430068, China.
Sensors (Basel, Switzerland)
|November 25, 2023
Summary
This study introduces a new multimodal framework using CNN-BiLSTM for early neurological disease detection through gait analysis. The method achieves high accuracy in identifying conditions like Parkinson's disease, aiding in early intervention.
Area of Science:
- Neurology
- Biomedical Engineering
- Data Science
Background:
- Global aging population increases neurological disease prevalence.
- Quantitative gait analysis offers early disease detection potential.
- Current single-sensor gait recognition methods face limitations in sensor type and task specificity.
Purpose of the Study:
- To develop a versatile multimodal gait-abnormality-recognition framework.
- To overcome challenges of data interference and long time series in gait analysis.
- To enable accurate identification of various neurological gait abnormalities.
Main Methods:
- Proposed a Convolutional Neural Network-Bidirectional Long Short-Term Memory (CNN-BiLSTM) framework.
- Employed an adaptive sliding window technique and time-frequency plots for unified data representation.
- Utilized a pre-trained Deep Convolutional Neural Network (DCNN) for feature extraction and multi-sensor data fusion.
Main Results:
- Achieved 98.89% accuracy in classifying Parkinson's disease severity, outperforming DCLSTM.
- Demonstrated high recognition accuracies for Amyotrophic Lateral Sclerosis (ALS) at 100%, Parkinson's disease (PD) at 96.97%, and Huntington's disease (HD) at 95.43%.
- The framework showed superior performance compared to most existing methods.
Conclusions:
- The multimodal CNN-BiLSTM framework shows significant potential for gait abnormality identification.
- Its adaptability to various sensors and fewer training parameters make it suitable for daily monitoring and personalized rehabilitation.
- This approach can help mitigate the impact of neurological diseases through early detection and tailored interventions.
Keywords:
CNN-BiLSTM networkgait abnormality recognitionmulti-sensorneurodegenerative diseasestime–frequency plots
