Human body activity recognition system and method based on conditional variational auto-encoder

Abstract

The invention discloses a human body activity identification system and method based on a conditional variational auto-encoder, and the method comprises the steps: obtaining an original time sequence:obtaining sampling samples through a sensor, and enabling a plurality of sampling samples to form the original time sequence; constructing batch data: constructing the batch data in a data enhancement mode of a random sequence starting point to obtain constructed sensor batch data X and corresponding activity label batch data Y; training a condition variation auto-encoder model: inputting batch data into the model, and training the model through a loss function and a back propagation algorithm; predicting human body activities: inputting the sensor batch data X as test data into the trained variational auto-encoder model, and inputting the batch data into the variational auto-encoder model to obtain a final prediction activity tag. According to the method, one sampling sample is used as aunit to predict the corresponding activity label, the real-time activity recognition capability is achieved, modeling can be conducted on the correlation of similar samples, and therefore the recognition accuracy is improved.

Description

technical field [0001] The present invention relates to human body activity recognition based on wearable devices, focusing on human body activity recognition based on a prediction unit of a sampling sample, and in particular to a human body activity recognition system and method based on a conditional variational autoencoder. Background technique [0002] Human Activity Recognition (HAR), also known as wearable sensor-based activity recognition, in this problem, we need time series triggered by sensors worn on the human body (such as three-axis accelerometer and heart rate meter) , to identify the current activity or posture of the person. Activity recognition has been one of the fundamental problems for many applications, such as fall detection, gesture recognition, etc. All in all, human activity recognition plays a very important role in many fields such as ubiquitous computing, intelligent care and behavior analysis. [0003] The general framework of human activity re...

AI Technical Summary

Problems solved by technology

However, there are some problems with this preprocessing. First, the sensor sampling samples contained in the subsequence do not necessarily all have the same activity label, and the subsequence may contain samples of two or more types of activities. At the same time, in practical applications, the optimal The selection of the window size is also a difficult problem

On the other hand, although deep models such as convolutional neural networks have achieved promising results, there are still many unsolved problems

For example, existing deep models cannot explicitly model the correlation of samples of the same class, i.e., cannot fully mine the inter-class similarity of activities

And this also hinders the further improvement of activity recognition accuracy.

Images