Dynamic gesture interaction method based on monocular camera

Abstract

The invention discloses a dynamic gesture interaction method based on a monocular camera, and the method comprises the steps: obtaining a plurality of video sequences containing different gesture actions, carrying out the normalization and preprocessing, and adding a label, so as to form a gesture interaction data set; obtaining a plurality of deep neural networks with the same structure to form aplurality of spatial feature extraction sub-branches; training a plurality of spatial feature extraction sub-branches based on the gesture interaction data set until convergence; obtaining a single deep neural network to form a time feature extraction total branch, and cascading the time feature extraction total branch with a plurality of spatial feature extraction sub-branches to obtain a gesture interaction model; adopting the gesture interaction data set to train the gesture recognition total model until convergence; and inputting a to-be-identified gesture video sequence into the converged gesture identification total model to obtain a judgment result. According to the invention, the difficulty of video sequence processing is reduced, and the precision level is ensured. The performance precision of the spatial feature extraction sub-branches is effectively ensured. Interactive identification of dynamic gestures can be realized.

Description

technical field [0001] The invention relates to the technical field of human-computer interaction, in particular to a dynamic gesture interaction method based on a monocular camera. Background technique [0002] With the development of science and technology, human-computer interaction has become a very important means of communication. Common human-computer interaction modes such as touch interaction, keyboard interaction, mouse control interaction, voice interaction, etc. Advances in computer vision are making gesture recognition a new mode of interaction. Compared with other methods, gesture recognition is more natural, simple and convenient, and can be used in applications such as smart home control and somatosensory games. Among them, the gesture recognition technology based on a monocular camera can be applied to ordinary color cameras, and the cost is low. Compared with expensive depth cameras, it is more suitable for large-scale and extensive deployment. [0003] ...

AI Technical Summary

Problems solved by technology

However, the detection of palms relies on traditional manual features, and the effect is difficult to guarantee, and the subsequent time recurrent neural network (LSTM) has a complex structure and is difficult to train, which brings difficulties to practical applications.

[0006] Therefore, the dynamic gesture interaction technology in the prior art has the technical problems of low precision and poor effect

[0007] For the above problems, no effective solution has been proposed

Images