Human body posture recognition method based on time and channel double attention

Abstract

The invention discloses a human body posture recognition method based on time and channel double attention. The method comprises the following steps: acquiring original data of various human body actions by using a built-in sensor of mobile equipment, attaching an attribute label of the action, performing sliding window and normalization processing, segmenting into a training sample set and a testsample set, establishing a deep convolutional neural network model based on time and channel double attention, and importing a training sample and a test sample for training and optimal adjustment toobtain a recognition result of the human body action. According to the invention, channel attention and time sequence attention are superposed, the type and the occurrence time of the target action can be accurately positioned after a large amount of coarse-grained training data is trained, so that the complexity of manually marking the training data is greatly reduced, and the invention plays animportant role in sports, interactive games, medical care, universal monitoring systems and the like.

Description

technical field [0001] The invention belongs to the field of intelligent monitoring of wearable devices, and in particular relates to a human body gesture recognition method based on time and channel dual attention. Background technique [0002] In recent years, with the development of computer technology and the popularization of smart technology, we have entered a new round of global technological change, and technologies such as large-scale cloud computing, Internet of Things, big data and artificial intelligence are also developing rapidly. Among them, human gesture recognition technology is also an important research trend in computer vision related fields. It has a wide range of applications and can be used in various fields such as health monitoring, motion detection, human-computer interaction, film and television production, and game entertainment. People can use the sensors worn by the human body to collect the motion trajectory data of human joint points to reali...

AI Technical Summary

Problems solved by technology

Compared with traditional machine learning methods, such as logistic regression, decision tree, Markov model, etc., although these shallow deep learning methods have significantly improved accuracy, the feature map information is not sufficient due to the small number of convolutional layers. Rich

At the same time, the pervasive convolution calculation cannot accurately locate the time when the target action occurs and accurately locate the type of target action in a series of long data.

Images