Attention mechanism-based LSTM human body behavior identification method

Abstract

The invention particularly relates to an attention mechanism-based LSTM human body behavior identification method. Human body joint point coordinate information is used as input data; human body jointpoints are divided into five groups according to a human body structure; the coordinate information is sent to five LSTM-Attention modules to carry out feature extraction; three times of local fusionis performed on the extracted new features, the new features are sent to a corresponding LSTM-Attention module for feature extraction to obtain whole human body features, and the whole human body features are sent to a full connection layer and a softmax layer to output a human body behavior recognition result. According to the method, an attention mechanism is introduced into the LSTM, so that the LSTM can well reserve and process time sequence information in data, the feature vector is transmitted into the Attention layer to adaptively perceive the network weight which has great influence on the recognition result, and the efficiency and accuracy of human body behavior recognition are improved.

Description

technical field [0001] The invention belongs to the technical field of computer vision, and in particular relates to an LSTM human behavior recognition method based on an attention mechanism. Background technique [0002] With the advent of the era of artificial intelligence, the realization of intelligence in various fields has become the general trend, and it will also bring great convenience to people's lives. As an important branch of artificial intelligence, computer vision can be regarded as the "eyes" of artificial intelligence. Its main task is to use computers to analyze and process the collected information (pictures or videos) to understand the semantic information contained therein. Human action recognition, as one of the most popular research directions in the field of computer vision, has attracted extensive attention from academia and business circles, and has broad application prospects in many fields in real life. [0003] The main purpose of human action r...

AI Technical Summary

Problems solved by technology

Traditional action recognition methods mainly use artificially designed features for feature extraction, but artificially designed features often need to be specifically designed according to different tasks, relying on the database itself, so its generalization ability and versatility are poor

In addition, traditional action recognition is mainly aimed at small data sets with small data volume, fewer types of actions, and lower action complexity.

However, under the background of today's information explosion and big data, the exponential growth of image and video data also makes the traditional method of action recognition based on artificially designed features unable to meet the demand

Images