Gait data de-noising based on improved EMD

The recovery of gait signal from observed noisy data is very important and classical problem in gait signal processing. Classical method such as Fourier transform and wavelet has some drawbacks when processing non-linear and non-stationary data like gait data, This paper describe a new method for gait accelerometer data de-noising base on EMD, a new envelop algorithm using Gaussian process is propose to improve the performance of EMD. The new algorithm is superior to existing classical algorithm because in most situations Gaussian process is more flexible than cubic spline interpolation algorithm. The method is fully data driven, and decomposes the signals in spatial domain; therefore it can discriminate the signals form the noise and could be used in both non-linear and non-stationary signals. New algorithm is used to de-noise gait data, the result are expected to show that it is better suited in de-noising gait data.