Dynamic gesture trace recognition method based on depth convolution neural network

Abstract

The invention provides a dynamic gesture trace recognition method based on a depth convolution neural network. The method comprises the following steps of collecting an originally input gesture trace point sequence; performing preprocessing ; detecting and eliminating abnormal points of the gesture trace point sequence; performing marginalizing processing on the preprocessed gesture trace point sequence; generating a normalized gesture trace graph; extracting depth features of the normalized gesture trace graph by using a trained depth convolution neural network model; recognizing the shape type of the corresponding gesture trace point sequence by using a trained support vector machine; dividing unknown direction types by a tree classifier according to the shape type of the gesture trace point sequence; fusing the recognized shape type and direction type; and generating a fused trace recognition result of the gesture trace point sequence. The method has the advantages that the shape recognition and the direction recognition are realized; the directed dynamic gesture recognition service is provided for the gesture trace point sequence; the dynamic gesture trace recognition work is not influenced by time and space differences; and the classification is fine.

Description

technical field [0001] The invention relates to the technical fields of computer vision and pattern recognition, in particular to a dynamic gesture track recognition method based on a deep convolutional neural network. Background technique [0002] With the continuous emergence of new technologies of artificial intelligence and new technologies of input and output equipment, human-computer interaction technology is rapidly moving towards the direction of intelligent automation, from the original computer-centered mechanical interaction technology to human-centered multi-channel multimedia intelligence Interactive technology comes up. These new human-computer interaction technologies get rid of the shackles of the old mechanical interaction, and are more and more popular among the audience, such as skin displays, fingerprint or corneal recognition security protection, eye movement interaction devices, etc. [0003] Gesture, as a common communication method for people, is a n...

AI Technical Summary

Problems solved by technology

However, this scheme is still based on the hidden Markov model, and is only interested in the state-space relationship composed of the trajectory point sequence, resulting in multiple cases being classified into the same type, and the recognition results are rough and not accurate enough.

Images