Objectives Aiming at the denoising problem of acceleration signal in gait recognition method, a parameter optimization method using composite evaluation index and wavelet entropy for wavelet denoising of gait acceleration signal is proposed.
Methods Firstly, based on the geometric and physical meanings of the four traditional evaluation indexes of wavelet denoising effect, namely root mean square error (RMSE), signal-to-noise ratio(SNR), correlation coefficient (R), smoothness (r), an improved entropy weight method is used to evaluate the denoising effect. The two indexes under different wavelet decomposition levels are linearly weighted to obtain a new comprehensive evaluation index. The decomposition level corresponding to the lowest point of the comprehensive evaluation index is the optimal decomposition level. Then, based on the physical meaning of wavelet entropy, the optimal wavelet base for wavelet decomposition is determined by calculating the wavelet entropy of the low-frequency coefficients of different wavelet bases to process the signal.
Results Simulation and test results show that: (1) Compared with the existing composite evaluation indexes, the composite evaluation index has certain advantages in reliability, simplicity and accuracy; (2) wavelet entropy based wavelet selection scheme (WEBWSS). takes into account the optimization of the number of decomposition layers and the wavelet basis, and has better denoising performance than WEBWSS; (3) The discriminant function method after optimization can locate the singularity of the signal well, but at the cost of the smoothness of the denoising signal. Under the premise of meeting the filtering requirements of gait signal research, the detailed features of the wavelet denoised acceleration signal waveform obtained by the WEBWSS are well preserved, and the gait cycle can be clearly distinguished.
Conclusions The wavelet denoising scheme determined by optimized WEBWSS can obtain satisfactory results when processing gait acceleration signals.
DownLoad: