Ankle Accelerometer FOG Detection Using Empirical Wavelet Denoising
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional gait disorder detection methods rely on patient-dependent models, which are cumbersome to construct and ineffective for sporadic freezing of gait (FOG) episodes, especially in home environments, and lack a practical, patient-independent model for wearable sensors.
Innovation Solution
A patient-independent model for freezing of gait detection using empirical wavelet decomposition, employing an ankle-worn accelerometer to process motion data through windowing, peak detection, and empirical wavelet decomposition to extract top-ranked features for classification into FOG and non-FOG classes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If patient-dependent models are used for FOG detection, then detection accuracy for individual patients may be improved, but model construction complexity and time consumption increase significantly
Solution Approach 1:
The patent develops a patient-independent FOG detection model that can universally detect FOG across different patients without requiring individualized model construction. The model uses empirical wavelet decomposition and machine learning algorithms to analyze gait patterns from wearable sensors, achieving accurate FOG detection while eliminating the need for patient-specific training and calibration procedures.
2Reliability
If multiple sensors are used to capture comprehensive gait data, then detection reliability is improved, but patient comfort and ease of operation deteriorate
Solution Approach 1:
The patent extracts and utilizes only the essential gait parameters needed for FOG detection from wearable sensor data, focusing on key features from accelerometer and gyroscope measurements. By identifying and extracting the most discriminative features related to FOG events, the system achieves reliable detection while minimizing the sensor burden on patients.
3Measurement precision
If complex signal processing techniques are applied to denoise motion data, then measurement precision is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent employs empirical wavelet decomposition to transform the motion data into a different representation domain, where noise can be more effectively separated from useful gait signals. By changing the parameter space through wavelet transformation and selecting appropriate decomposition levels, the system achieves effective denoising while maintaining computational efficiency for real-time FOG detection.
Data Source
AI summary
This disclosure relates generally to patient invariant model for freezing of gait detection based on empirical wavelet decomposition. The method receives a motion data from an accelerometer sensor coupled to an ankle of a subject. The motion data is further processed to denoise a plurality of data windows using a peak detection technique to classify into a real motion data window or a noisy data window. Further, a plurality of denoised data windows are generated by processing spectrums associated with each real motion data window and a plurality of empirical modes using an empirical wavelet decomposition technique (EWT). Then, a resultant acceleration is computed, and a plurality of features are extracted from the denoised data window which enables detection of freezing of gait based on a pretrained classifier model into a (i) a positive class, or (ii) a negative class.