Multichannel electromyographic signal processing method based on noise-assisted multivariate empirical mode decomposition

Abstract

The invention discloses a multichannel electromyographic signal processing method based on noise-assisted multivariate empirical mode decomposition. The method comprises the following steps: synchronously collecting exciting bioelectric signals generated by motor nerves at muscle groups when multiple muscle groups generate movement of a limb, obtaining a multichannel electromyographic signal (MEMG), constructing a multichannel white noise sequence having the same data length as an input signal by means of the properties of EMD, EEMD and MEMD structures, superposing the multichannel white noise sequence with a picked up MEMG signal, mapping data into a multi-dimensional vector space through a low-difference Hammersley sequence, and then adaptively decomposing the data into a plurality of intrinsic mode functions in a multi-variable condition based on the empirical mode decomposition. With respect to mode queue, the mode queue of the MEMD and the NA-MEMD are higher than that of the EEMD, however the improvement effects of the NA-MEMD and the MEMD to the mode queue are close on the whole; and with respect to mode aliasing, the frequency overlapping coefficient of the NA-MEMD is smaller than that of the MEMD, and the frequency overlapping coefficient of the MEMD is smaller than that of the EMD.

Description

technical field [0001] The invention belongs to the field of electric signal processing, in particular to the technical field of synchronous analysis of multi-channel or multi-variable non-stationary and non-linear myoelectric signals. [0002] technical background [0003] Electromyography (EMG) is a representation of the physiological activities of the human neuromuscular system on the skin surface of the body during muscle contraction, and is a complex nonlinear and non-stationary physiological electrical signal. Since SEMG is a direct representation of the body's neuromuscular system, mining SEMG signals, especially the physiological mechanism of multi-channel EMG (Multiple-channel EMG, MEMG), can reveal the functional characteristics of the body's neuromuscular system to a certain extent. In terms of physiology, it can provide a theoretical basis for the study of the physiological feedback mechanism of the human neuromuscular system; in clinical medicine, it can also pro...

AI Technical Summary

Problems solved by technology

(1) Wavelet analysis, this method makes the time-frequency characteristics of the decomposed signal components have better independence through continuous differentiation with the help of wavelet basis functions, and has advantages in real-time processing of SEMG signals, but because wavelet decomposition has cross terms, It cannot fully describe the natural properties of multi-component and non-stationary SEMG signals, let alone explain the characteristics of motor neurons reasonably

(2) High-order statistics. This method models the SEMG signal according to the linear time-invariant model, which can better evaluate the degree of nonlinearity, non-stationarity and non-Gaussianity of the muscle constant force contraction (static) SEMG signal, but The disadvantage is that this method cannot analyze the SEMG characteristics under the action of dynamic load or muscle force

(3) Artificial neural network. Although this method can describe SEMG signals well, it also has certain limitations. For example, the training process of artificial neural network often requires a large amount of prior data.

(4) Independent component analysis. This method assumes that the noise information and the motor neuron information belong to different source signals. By constructing a signal mixture model for inverse calculation, extracting signal patterns from complex signals, it can handle non-Gaussian signals, but The component pattern obtained by this method is lost, and it is often impossible to reversely reconstruct the original signal

Images