Convolutional neural network-based human body behavior recognition method and recognition system

Abstract

The invention proposes a convolutional neural network-based human body behavior recognition method. The method comprises three stages, namely, a data acquisition and preprocessing stage, a model establishment and training stage and a model application stage. According to the method of the invention, an established convolutional neural network structure can effectively identify six kinds of daily human behaviors on the basis of simple data preprocessing and provide a good foundation for transplanting the recognition method into a smart phone with limited computing power. The method can be applied to application fields such as smart phone end-based human body daily activity recording, health tracking and health monitoring.

Description

technical field [0001] The invention relates to six kinds of human daily behaviors, including walking, jogging, going up and down stairs, standing and sitting, state analysis and health monitoring, etc., specifically a human behavior recognition method and recognition system based on a convolutional neural network. Background technique [0002] With the development of information technology, smart phones integrating various sensors (such as three-axis acceleration sensors and gyroscopes, etc.) are becoming more and more popular. In recent years, more and more application software based on human behavior recognition based on mobile phone sensors have been developed to record and monitor the daily behavior of the human body, and at the same time can effectively help users establish a healthy living habit. In this field, seeking An efficient identification method of human physiological activities has become the most critical, core and urgent problem. [0003] One of the common...

AI Technical Summary

Problems solved by technology

[0003] One of the common smartphone-based human behavior recognition methods is implemented by using the Google Activity Recognition API (Google Activity Recognition API), which can recognize human behaviors such as cycling, running, walking, and stillness. , but the accuracy rate of human behavior recognition realized through this interface is very low, the reason is that the user's motion patterns (movement speed, gait, etc.) For people of gender, the relatively single and fixed methods of judging behavioral activities are not all applicable and effective

Images