Human body behavior identification method based on Bi-LSTM-Attention model

Abstract

The invention provides a human body behavior identification method based on a Bi-LSTM-Attention model. The method comprises the following steps: step S1, inputting an extracted video frame into an Inception V3 model, using the Inception V3 model to increase the depth of a convolutional neural network and reduce network parameters at the same time, and fully extracting depth features of the video frame to obtain related feature vectors; step S2, transmitting the feature vector obtained in the step S1 to Bi-LSTM neural network for processing, fully learning time sequence characteristics betweenthe video frames through the BI-LSTM neural network; and step S3, transmitting the time sequence feature vectors obtained in the step S2 to an attention mechanism model to adaptively perceive the network weight which has a great influence on the identification result, so that the features related to the network weight can get more attention. The recognition rate of human body behaviors can be improved.

Description

technical field [0001] The invention relates to the field of video analysis and recognition, in particular to a human behavior recognition method based on a Bi-LSTM-Attention model. Background technique [0002] For human behavior recognition, in the early days, most of them used artificially designed methods to extract video features. One scheme uses the method of spatio-temporal interest points to extract human body features in a complex background. This method calculates the strength of each position in the video sequence and finds spatio-temporal interest points through the method of maximum value filtering. WANG W et al. use the method of sparse coding to learn static features, and use the time-domain pyramid structure based on maximum pooling to represent the features in a histogram, and finally use SVM for classification. Another scheme proposes a hierarchical clustering multi-task learning (HC-MTL) approach to realize human action recognition by enforcing shared act...

AI Technical Summary

Problems solved by technology

However, the depth of CNN has a great influence on the feature extraction of video frames: the low network level is not easy to show the deep features of the image, and it is easy to underfit; the deep network model is prone to gradient dissipation and it is difficult to optimize the network model.

LSTM cannot effectively learn the temporal characteristics of motion and lacks the ability to adapt autonomously

Images