Gait episode identification based on wavelet feature clustering of spectrogram images

Measurement of gait parameters can provide important information about a person's health and safety. Automatic analysis of gait using kinematic sensors is a newly emerging area of research. We propose a new approach to detect gait episodes using Neural Network and and clustering of wavelet-decomposed spectrogram images. Signals from a chest-worn inertial measurement unit (IMU) is processed using Explicit Complementary Filter (ECF) to estimate and track torso angle. Using the feature obtained from wavelet decomposition of spectrogram images, we use an Augmented Radial Basis Neural Network (ARBF) to classify walking episodes. Cluster centroids of ARBF are optimized using Rapid Cluster Estimation (RCE). A pilot study of 11 participants suggests that our approach is able to distinguish between walk and non-walk activities with up to 85.71% sensitivity and 91.34% specificity.