Quantifying Stability Using Frequency Domain Data from Wireless Inertial Measurement Units

Abstract

The quantification of gait stability can provide valuable information when evaluating subjects for age related and neuromuscular disease changes. Using tri-axial inertial measurement units (IMU) for acceleration and rotational data provide a non-linear profile for this type of movement. As subjects traverse various surfaces representing decreasing stability, the different phasing of gait data make comparisons difficult. By converting from time to frequency domain data, the phase effects can be ignored, allowing for significant correlations. In this study, 12 subjects provided gait information over various surfaces while wearing an IMU. Instabilities were determined by comparing frequency domain data over less stable surfaces to frequency domain data of neural network (NN) models representing the normal gait for any given participant. Time dependent data from 2 axes of acceleration and 2 axes of rotation were converted using a discrete Fourier transform (FFT) algorithm. The data over less stable surfaces were compared to the normal gait NN model by averaging the Pearson product moment correlation (r) values. This provided a method to quantify the decreased stability. Data showed progressively decreasing correlation coefficient values as subjects encountered progressively less stable surface environments. This methodology has allowed for the quantification of instability in gait situations for application in real-time fall prevention situations.

Keywords

• Neural networks
• Biomedical Informatics
• stability
• Fourier Transform
• Biomechanics