Human Complex Behavior Recognition Method Based on Multi-Feature Fusion CNN-BLSTM

Abstract

A complex human behavior recognition method based on multi-feature fusion CNN-BLSTM includes the following steps: segment continuous sensor data through a sliding window, extract features from the segmented sensor data, and use a feature selection algorithm to classify the series of artificial Extract features for screening and retain dominant features; input the segmented behavior data into the deep learning model for training, first perform one-dimensional convolution pooling processing through the convolutional neural network, and then use the average pooling process through the bidirectional long-term and short-term memory neural network The layer extracts the salient features of the state output by the bidirectional long-short-term memory neural network, and finally fuses the pooled feature vector with the previously extracted advantage feature vector as the input feature of the fully connected layer to obtain the output of complex behavior recognition. The invention can fully excavate the characteristics of sensor data and improve the recognition accuracy of complex human behaviors.

Description

technical field [0001] The present invention relates to the fields of data analysis, feature extraction, deep learning, recurrent neural network, behavior recognition, etc., especially to a method for human complex behavior recognition Background technique [0002] Human complex behavior recognition has become a prominent research field at present. The main purpose of complex behavior recognition is to conduct continuous and real-time behavior observation on the behavior occurrence objects. In recent years, the rapid development of the Internet of Things industry, the rapid popularization of smart phones, bracelets, watches, etc., is accompanied by the rapid development and application of various sensors. Compared with the human behavior application of video detection, the sensor-based human behavior recognition method is low in cost, convenient, and has good portability. At present, rich sensing devices have been integrated on smart terminals, such as acceleration sensors,...

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the problems of manual extraction of features by traditional machine learning methods and incomplete feature extraction by a single deep learning model, as well as the problems of gradient dissipation and gradient explosion in the process of backpropagation of recurrent neural networks, the present invention proposes a method based on one-dimensional convolutional neural network Network and bidirectional long-short-term memory neural network (CNN-BLSTM) human body complex behavior recognition method, this method combines manual extraction features and deep learning model features, can overcome the traditional machine learning and the current single deep learning model cannot fully extract time series data features problems, and overcome the long-term dependence of the recurrent neural network, and improve the accuracy of human complex behavior recognition

Images