A fall risk assessment method based on wearable sensors
By combining data from inertial sensors and plantar pressure sensors using the ISA-ECA-CNN-BiLSTM network model, adaptive weighted feature fusion is performed, which solves the problem of insufficient robustness of single-modality sensor signals and achieves highly accurate and fast fall risk assessment.
CN116491933BActive Publication Date: 2026-06-19HEBEI AGRICULTURAL UNIV.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HEBEI AGRICULTURAL UNIV.
- Filing Date
- 2023-04-26
- Publication Date
- 2026-06-19
AI Technical Summary
Technical Problem
Existing fall risk assessment models based on single-modal sensors are not robust enough and lack effectiveness, and most models lack real-world everyday data.
Method used
Feature extraction is performed using an ISA-ECA-CNN-BiLSTM network model. Combined with data from inertial sensors and plantar pressure sensors, fall risk detection is performed through adaptive weighted feature fusion. Sequence samples are constructed using a sliding window method, and an improved ECA attention module and spatial attention mechanism are added.
Benefits of technology
It improves the accuracy and speed of fall risk assessment, enhances the stability and recognition ability of the model, and solves the problem of insufficient robustness of single-modal signals.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116491933B_ABST
Abstract
This invention discloses a fall risk assessment method based on wearable sensors, comprising the following steps: S1, data is collected using inertial sensors and plantar pressure sensors through a human walking experiment, and labeled using the TUGT test, Tinetti balance, and gait scale. The collected data is filtered and segmented to obtain a gait sequence dataset; S2, the gait sequence dataset is input into an ISA-ECA-CNN-BiLSTM network model for feature extraction; S3, the features extracted from the IMU signal and plantar pressure signal are adaptively weighted and fused to perform a fall risk detection task. This invention employs the above-mentioned fall risk assessment method based on wearable sensors, solving the problems of insufficient robustness of single-modality sensor signals and lack of effectiveness in risk assessment models. By using the collected gait dataset to learn gait features, this method has advantages such as high accuracy, fast recognition speed, and stable performance.
Need to check novelty before this filing date? Find Prior Art