Intent recognition of torso motion using wavelet transform feature extraction and linear discriminant analysis ensemble classification

In this paper, a multi-sensor, multi-classifier approach for intent recognition of human torso motion is presented. A linear discriminant analysis based classifier is used, and the extraction of time-frequency domain features through the use of the wavelet transform is discussed. In addition, a weighted multi-classifier combination method for combining outputs of multiple classifiers into a single coherent output is implemented. The approach was evaluated on physiological data collected from three human participants. Results show up to 97% accuracy in classifying flexion and extension motions of the torso.