Motion intention recognition method and device based on limb morphological characteristics

Abstract

The invention provides an action intention recognition method and device based on limb morphological characteristics. The method comprises the steps of receiving bending degrees of a cross section contour of a to-be-detected limb in multiple directions; wherein the bending degrees of the cross section contour in multiple directions are acquired by a wearable sensing device worn on the limb to be detected, and the wearable sensing device comprises a plurality of bending degree sensors which are uniformly distributed in the circumferential direction; fitting a closed curve of the cross section contour based on the bending degrees of the cross section contour in multiple directions, and determining morphological characteristics of the cross section contour based on a fitting result of the closed curve of the cross section contour; and inputting the morphological features of the cross section contour into a pre-trained intention recognition network to obtain an action intention recognition result of the to-be-detected limb. The method can effectively ensure that the obtained action intention recognition result is earlier than the actual motion, and overcomes the limitation of the traditional action intention recognition method on time.

Description

technical field [0001] The present invention relates to the technical field of human-machine interfaces, and in particular, to a method and device for recognizing action intentions based on morphological features of limbs. Background technique [0002] Wearable human-machine interface is a carrier to realize human motion and recognition, and plays an important role in human-computer interaction and wearable robot control. Human-machine interfaces include wearable sensing systems and intent-recognition algorithms. Among them, the sensing system is the first key link to extract the wearer's motion characteristics, and its performance directly determines the effect of the subsequent recognition algorithm and the final recognition performance. Based on the prior art in the field, the signal sources that the sensing system can extract include biological signals, kinematic signals and interactive force signals. Among them, biological signals mainly include signals that can be ex...

AI Technical Summary

Problems solved by technology

[0004] However, most of the existing fabric sensors that measure stretch and pressure mechanical signals measure the movement of joints or the information of interaction with the outside world. From the perspective of movement characteristics, they belong to the response to movement, that is, movement occurs first and then generation The signal characteristics change, and the collected signal lags behind the actual movement in time. Therefore, there are time limitations in the process of action intention recognition

Images