Motion Intent Recognition in Intelligent Lower Limb Prosthesis Using One-Dimensional Dual-Tree Complex Wavelet Transforms

Algorithm 1

The proposed motion intent recognition.

	Input: discrete motion behavior data collected by inertial sensors (three-axis data of angular velocity and acceleration of thigh, shank, and foot).
	Output: recognition accuracy of motion intent
(1)	Use moving average filter to process collected discrete motion behavior data
(2)	The processed data are decomposed by 1D-DTCWT
(3)	Select -layer of low-frequency coefficients from 1D-DTCWT as the continuous features, forming 72 (34) dimensional feature vector
(4)	Use SVM for motion states classification
(5)	Adopt the confidence level to analyze the experimental results
(6)	return recognition accuracy of motion intent